O F TH E VICTORIA, B.C. SKIER M ARKET
by
Barbara Anne Carmichael
B.A. University of Bristol, U.K., 1969 M.Sc. University o f Durham, U.K., 1.980
A Dissertation Submitted in Partial Fulfilment of the o a i |. ^ C E P T E D Requirements for the Degree of
n CULTY
OFn R A h i i A T r
q T . ._C ^ f U D f E S DOCTOR O F PHILOSOPHY
O A T fr / ! I ^ DEAN
- - ii / We accept this dissertation as conforming in the Department of Geography
:pt this dissertation as con to the required standard
Dr. P.E. Murphy, Supervisor (JZiej^rtmGftt ofy^ography)
Dr. C. Wood, bepafw^etflaTM ember (Department of Geography)
Dr. C. Forward,"Dcpartme.nt Member (Department of Geography)
Dr. R. Gifford, Outside Mefnber (Department of Psychology)
— - ' |— — ...
Dr. P. Williams, External Examiner (Centre of Tourism Policy and Research, School of Resource & Environmental Management, S.F.U.)
©BARBARA ANNE CARMICHAEL, 1991 University of Victoria
All rights reserved, Dissertation may not be reproduced in whole or in part, by photocopying or other means, without the permission of the author.
Supervisor; Dr. Peter E. Murphy
ABSTRACT
Previous studies in travel and tourism, both in the tourist industry and within geography have highlighted the need for perceptual studies linking the formation of tourist image and evaluation of tourist resources to destination choice. This study sets out to explore the processes of consumer image formation, decision-making and destination choice, mainly within the context of the B.C. domestic ski market, but with wider implications for travel choice decision-making in general.
A conceptual and methodological framework is developed in which tourist images are measured using multi-attribute, conjoint and hybrid models. These indices of image are used in predictive choice models to assess the performance of perceptual variables as compared with objective variables, in the prediction of actual skier choice behaviour. This framework is applied to 359 skiers, who are residents of Victoria, B.C. and who are in the anticipation stage of choosing a skiing holiday. They are contacted by pre-trip and post-trip telephone interviews and, in addition, a subgroup of 100 are studied through personal interviews.
The findings of this study reveal that the models which use perceptual data markedly outperform in predictive ability those using objective data. All models using perceptual data achieve good results both at the individual and aggregate levels. The study also reveals substantive findings of benefit to the ski industry. The relative performance of resorts for the Victoria ski market, as found by comparing their image scores, and the relative importance of ski resort attributes in their contribution to resort choice, as revealed by conjoint analysis, have useful implications for s . ; . “.sort
marketing. These results of both a theoretical and substantive nature provide suggestions for future research and planning within the ski industry.
Examiners:
Dr. P.E. Murphy, Supervis6r (Dcparfmenj^f'Dcography)
Dr. C. W ooifD^pSttrtGntal Member (Department of Geography)
Dr. C. Forward, foeptfrtment Member (Department of Geography)
Dr. R. Gifford, Outside Member (Department of Psychology)
Dr. P. Williams, External Examiner (Centre of Tourism Policy and Research, School of Resource & Environmental Management, S.F.U.)
iv TABLE O F CONTENTS
Page
ABSTRACT ... ii
TABLE O F C O N TEN TS... iv
LIST OF TA B LES... vii
LIST OF FIGURES ... viii
ACKN O W LED G EM EN TS... ix
DEDICATION ... . ... x
CHAPTER 1 - IN TR O D U C TIO N ... 1
1.1 Introduction . . ... ... 1
1.2 Geography and Tourism S tu d ie s ... 2
1.3 Geography and Behavioral Studies o f Image and Decision-Making ... 3
1.4 Tourism and Behavioral Studies in Tourist Image and Decision-Making . . . . 6
1.5 Sample S t u d y ... 12
1.6 Research H ypothesis... 12
1.7 Dissertation O u tlin e ... 13
CHAPTER 2 - THEORETICAL AND METHODOLOGICAL F R A M E W O R K ... 14
2.1 Approaches to Model B u ild in g ... 14
2.2 Approaches to Model Building in This R e se a rc h ... 17
2.2.1 Tourist Image M easurem ent... 18
2.2.2 Decision Rules ... 25
2.3 Research Objectives ... 26
CHAPTER 3 - RESEARCH D E S IG N ... 2,8 3.1 In tro d u c tio n ... 28
3.2 Preliminary Study: Attribute Generation ... 30
3.3 Pre-trip Telephone Survey... 39
3.4 Post-trip Telephone S u rv e y ... ... 45
3.5 Personal Interview s... 46
CHAPTER 4 - DATA ANALYSIS AND R E S U L T S ... 52
4.1 Introduction ... 52
4.2 Cltoice of Salient Attributes of Resort Attractiveness ... 52
4.2.1 Pre-Trip Questionnaire Results ... 53
Page
4.3 Tourist Image Measurement M o d e ls ... 55
4.3.1 Multi-attribute M odel... 56
4.3.2 Conjoint M o d e l ... 59
4.3.3 Hybrid M o d e l... 74
4.4 Prediction of Ski Resort Choice on an Individual B a s is ... 77
4.5 Actual Ski Resort C h o ic e ... 81
4.6 Prediction of Ski Resort Choice on an Aggregate B a s is ... 85
4.7 Geographical Position and Spatial P a tte rn s ... ... 88
4.7.1 Island Location ... 88
4.7.2 Coastal Mainland R e so rts ... 88
4.7.3 Inland B.C... 90
4.7.4 Resorts in Alberta and U.S.A... 90
4.8 Prediction of Ski Resort Choice Using Physical Resort A ttrib u te s ... 90
CHAPTER 5 - C O N C L U S IO N ... 94 5.1 Restatement of Purpose ... 94 5.2 Theoretical Is s u e s ... ... ... 94 5.3 Substantive Is s u e s ... 97 5.3.1 Resort Images ... 97 5.3.2 Resort A ttrib u te s ... 98 5.4 Methodological Lessons ... 99
5.4.1 The Pre-trip Telephone Interviews ... 99
5.4.2 Post-trip Interviews ... 99
5.4.3 The Personal In terview s... 99
5.5 Suggestions for Future R e se a rc h ... 100
B IB L IO G R A P H Y ... 102
APPENDIX A - Telephone Questionnaires ... 113
APPENDIX B - Computation Methods Using SPSS-PC+ ... 122
APPENDIX C - Kendall’s Coefficient of Concordance A Validity Test for the Importance of Attributes (Ranked from Conjoint Analysis Compared With Overt Ranking Test) ... 125
APPEND k D - Differences in Importance of Attributes Between the Short-trip and the Long-trip C ontexts... ... 126
APPENDIX E - Calculation of Aggregate Probabilities (Models la - 9 a ) ... 127
APPENDIX F - Calculation of Aggregate Probabilities (Models lb - 9b) ... 137
vi APPENDIX G - Calculation o f Aggregate Probabilities
(Models 10 -1 1 ) ... 147 APPENDIX H - Comparison of Actual Market Shares and
m 4 31 33 34 38 41 47 54 58 61 62 64 73 76 79 80 82 87 91 LIST O F TABLES
Early Research Influencing Behavioral G eography... Academic Studies of Skier D em an d ... Percentage Number of Times Ski Facilities Mentioned ... Qualities of the Ski E xperience... Attributes of Ski Resorts ... Sample Size T a b le ... Card Profiles for Ski Resort Ranking T a s k ... Other Factors Important in Resort Choice
(Personal Interview D a t a ) ... ... Image Scores Computed Using Multi-attribute Models ... Subfile Summary Data From Conjoint Analysis
For 88 Skiers Considering a Short Ski Trip (1-3 days s k iin g )... Subfile Summary Data From Conjoint Analysis
For 82 Skiers Considering a Longer Ski Trip (4 or more days skiing) Summary Table of Importance Values
For Attributes in Different S itu a tio n s ... Image Scores Computed Using Conjoint M o d e ls... Image Scores Computed Using Hybrid Models ... Percentage Success Rates of Models at the Individual Level ... Percentage Success Rates for Total Sample
(Skiers considering more than one r e s o r t) ... Actual Ski Resort Choice and
Travel Behaviour (Total sample results) ... Correlation Between Actual and
Predicted Trips (n = 23 resorts v isited )... C ot.elation Coefficients Between Actual and Predicted Trips ...
viii LIST O F FIGURES
Page
Figure 1.1 - Stages in the Tourist Decision Process ... 10
Figure 1.2 - Conceptual Model of Image/Choice R elationships... 11
Figure 2.1 - A Framework for Geography and Marketing M o d e ls... 15
Figure 2.2 - Stages in Research Design ... 19
Figure 3.1 - Research D e sig n ... . . 29
Figure 3.2 - Reasons for Selecting a Ski Area ... 35
Figure 3.3 - Importance Performance Analysis of B.C. Ski Resorts ... 36
Figure 3.4 - Examples of Resort Profiles ... 50
Figure 3.5 - Attribute Curds for R a n k in g ... 51
Figure 4.1 - Image Measurement M o d e ls ... 55
Figure 4.2 - Group Average Part Worth Functions (1-3 days sk iing)... 65
Figure 4.3 - Group Average Part Worth Functions (4 or more days skiing) ... , ... 66
Figure 4.4 - Frequency of Pearson’s R in Model Fit for Conjoint Analysis (1-3 days skiing) ... 68
Figure 4.5 - Frequency of Pearson’s R in Model Fit for Conjoint Analysis (more than 4 days skiing) ... , ... 69
Figure 4.6 - Frequency of Pearson’s R Between Observed and Predicted Rank Order on Holdout Cards ... 70
Figure 4.7 - Frequency of Pearson’s R Between Observed and Predicted Rank Order on Holdout C a rd s ... 71
Figure 4.8 - Actual Travel Behaviour (post-trip survey r e s u lts ) ... 78
Figure 4.9 - Resorts in Skier’s "Evoked" Sets 1989-90 (Skiers from Victoria, BC) ... 83
Figure 4.10 - Resorts Actually Chosen by Skiers 1989-90 (Skiers from Victoria, BC) ... 84
Figure 4.11 - Geographical Position and Spatial Patterns of Behavior ... 89
Figure 5.1 - Conceptual Framework for Geographical Models of Travel Behaviour (based on Cadwallader, 1989, p. 496) ... 94
ACKNOWLEDGEMENTS
I would like to thank a number of people who provided help and support to me during my research and writing.
All of my Committee deserve special thanks for their time and helpful remarks and suggestions but I would especially like to thank my Supervisor, Dr. Peter Murphy, for his encouragement and support throughout my graduate work, He always gave freely of his time and read my various drafts promptly and carefully.
I am also grateful to the retail stores, ski clubs and skiers who participated im my research and without whom the research would not have been possible.
Finally, a very special thanks goes to my husband Bill and my teenage sons, Andrew and Simon, who coped really well while I was so preoccupied with my work and offered the moral support and confidence in me which I needed to get to this stage in my academic career.
DEDICATION
INTRODUCTION
1.1 Introduction
If wc arc going to influence a decision, a change in behaviour, we *,eed to know how that decision is made. There must be an increasing focus in the research and planning in our industry (travel and tourism) on how decisions arc made, It means wc need to know more than demographics. The terrain is in the buyer’s mind; not in just who he or she is. (Davidson, 1985, p. 106)
As Davidson states there is need for more research on the processes involved with tourist decision-making and destination choice. This research area is of scholarly interest to behavioral geographers as well as of being o f practical significance to the tourist industry. The central question to be addressed in this dissertation is the link between tourist images and perceptions and their destination choice. In this first chapter an overview is presented o f the geographical and tourism literature to outline the scholarly significance and justification of this research. First, the contribution of geography to tourism research is reviewed and emphasis placed on the currently expanding research frontier where geographers are developing their interests in consumer image, perceptions and decision-making and their marketing implications. Second, the origins of this work are traced to the contribution of behavioral geographers to the study of environmental perception, image and decision making in a wider geographical context. Third, reviews arc made in more detail of the contribution of tourist geographers and other researchers to <hc tourist image literature and to the advances in knowledge of destination choice.
The case study context of the current research is then specified. To test the hypothesis that there is a link between tourist image and destination choice, a sample study was made of a group of skiers from one destination (Victoria, B.C.), who were in the anticipation stage of considering a skiing holiday during the 1989-90 ski season,
2 1.2 Geography and Tourism Studies
Geographers are interested in the explanations, predictions and synthesis of spatial patterns and processes within the natural and cultural landscape. Pearce (1979, p. 248) defined tourism geography as being "concerned essentially...with the spatial expression of...the relationships and phenomena arising out of the journeys and temporary stays of people travelling primarily for leisure or recreation purposes." Mitchell (1987, p. 187) suggested that Pearce’s definition consists of five parts: (1) space or place; (2) relationships; (3) phenomena, facilities and activities; (4) travel; and (5) leisure or recreation. Therefore, tourism geographers observe, analyze and explain the relationships of specific activities and facilities that are located in particular tourism areas or regions (Mitchell, 1987, p. 193).
Mitchell and Murphy (1991) in their review of the geography of tourism outlined the key areas of focus in geographical tourism research. Much of the emphasis in the geography of tourism has been on travel patterns (Pearce & Elliot, 1983; Murphy & Keller, 1990), travel predictions (Baxter
& Ewing, 1986), tourist impacts (Mathieson & Wall, 1982), and patterns o f location and tourist space
(Yokeno, 1974; Miossec, 1977; Husbands, 1983). Geographers have also been interested in the environmental impacts of tourism (Kariel, 1988; Singh, 1989), and in sustainable development (Moser & Moser, 1986; Butler, 1989). Some research has focused on host communities and perceptions and attitudes towards tourism (Murphy, 1980) as well as on appropriate levels of decision-making and community involvement (Murphy, 1985).
As yet there has been little development of theoiy, although certain models borrowed from other disciplines have been introduced. Butler’s application of the product life cycle concept to tourist destinations (Butler, 1980) spawned many geographical follow-up studies (Lundgren, 1983; Keller, 1987; Foster, 1988; Strapp, 1988; Cooper, 1990). Economic impact models borrowed from economics (Archer, 1974) have also received attention (Liu & Var, 1981).
The research interests described so far have concentrated on the supply side of tourism. However, as Leiper (1979) suggested, tourism facilities should be regarded as part of a wider system
in which individuals, destinations and travel are interconnected. Similar ideas were expressed by Mill and Morrison (1985, p. 357) in their discussion of the tourist system. In addition, Lcipcr (1990) further developed his system’s framework by discussing a tourist attraction system, defined as an empirical connection between tourist, nucleus and markers. The nucleus is any feature or characteristic of place that a traveller contemplates visiting. Markers are items of information about any phenomenon that is a potential nuclear clement in a tourist attraction. It seems therefore that research which focuses on the linkage between tourist, place attributes and information could provide a broad based understanding of how a destination really attracts tourists.
Interestingly, some recent geographical research is emphasizing the demand side of tourism and exploring tourist choice behaviour. Recent tourism monographs published by geographers in the United Kingdom and in the Netherlands have focused more on tourist image and tourism marketing (Goodall & Ashworth, 1988; Ashworth & Goodall, 1990). Understanding holiday choice is a subject incorporated in a recent volume of "Progress in Tourism, Recreation and Hospitality Management" (Goodall, 1991), and a forthcoming book "Choice and Demand in Tourism" (Thomas & Johnston, awaiting publication) shows the current surge of academic interest in this subject area in the United Kingdom. In Canada, a new book by Wall and Heath is awaiting publication under the title of "Marketing of Tourist Destinations".
1.3 Geography and Behavioral Studies o f Image and Decision-Making
Although image and choice studies are relatively new in tourism, they are well established in geography, as behavioral geographers in the 1960s and 1970s have sought to study human behaviour in space and time. A review by Thrift (1981) outlined the many origins and directions of this behavioral research. Essentially, there were- three main roots to behavioral geography which followed parallel but separate strands of development. The three areas were environmental perception research, natural hazard research; and quantitative research on choice and decision-making. The major contributions to early work in these three areas are summarized in Table 1.1.
4
Table 1.1
Early Research Influencing Behavioral Geography
Environmental Perception Natural Hazards
Choice and Decision-Making Boulding 1956 (image) Kates 1962 (flood hazard) Huff 1960 (consumer behaviour) Lynch 1960
(image of the city)
Gould 1963 (management/environment relationships) Wolpert 1965 (migration decision/place utility) Lowcnthal 1967 (landscape perceptions) Saarinen 1969 (drought perception) Prcd 1967 (behavioral matrix)
Since the research directions followed by researchers in the three areas were very different, this led to subsequent difficulties in the definition of behavioral geography (Goodey & Gold, 1985). Some studies, especially in choice and decision-making followed a positivist analytical approach (Gollcdgc & Couclelis, 1984). They stressed the "importance of logico-mathematical thinking, the desire for public verifiability, the search for generalization and the emphasis on analytic language for expressing knowledge (Gollcdge & Couclelis, 1984, p. 180). Other studies rooted in environmental perception favoured a more descriptive phenomenological approach, being concerned with resurgent themes of place, landscape, dwelling, region and dialects (Mugerauer, 1984). Hazards research developed a path midway between the two. Geographical research in both environmental perception and in choice and decision-making pertains directly to this current dissertation research and in the rest of this section a review will be made of geographers’ contributions in these areas.
Environmental perception of place is the individual’s mental structuring of the physical and social environment. Pocock and Hudson (1978, p. 3) defined the role of environmental images as "information storage devices to classify, categorize and so differentiate and assign meanings to locations." Images are based on perceptions. This perception is a filtering process; a person selects what he will perceive based on what is important and relevant to him or her (Mayo & Jarvis, 1981). A person’s behaviour is based on his or her perception o f the environment, not on the environment as it actually exists. Kirh (1963) made the important distinction between the "phenomenal" or
objective physical environment in which behaviour takes place and the "behavioral" environment, which is the subjective psychological environment in which decisions arc made.
How far this subjective environmental perception may be measured is a point of debate in behavioral geography. In the present study, a quantitative analytical approach is adopted as the research interest is in model building and prediction of behaviour. However, much interesting work in humanistic behavioral geography has explored the ideas of image and perception using methodologies of description and participant observation.
Environmental perception studies caught the attention of geographers since they improved the understanding of man-environment relationships. Studies of regional images also provided a fresh perspective for regional geographers. Wood (1970, p. 129, cited in Pocock & Hudson, 1978) stressed that the "understanding of a vast range of materials which human geographers study can be greatly increased by a consideration of an individual’s perception of his environmcnt...thc study of perception may in lime, achieve a deep understanding of the man-environment relationship."
Humanistic geographers have developed the environmental perception research theme by exploring ideas on the uniqueness of place and the meaning of place io individuals. Yi-Fu Tuan (1977) described places as centres of perceived personal value. He developed the term "topophilia" to describe a human being’s affective tics witn his or her material environment and particularly with specific places and settings (Johnston, 1988, p. 493). The exact nature of such ties has varied enormously in intensity, subtlety and mode of expression; and the response itself may be primarily aesthetic, tactile, emotional, nostalgic or economic in determination.
Sense of place and also of placelessness was studied by Relph. Rclph (1976, p. 43) stated that "there is for virtually everyone a deep association and consciousness of the places where we were born and grew up and where we live now or where wc have had particularly moving experiences. This association seems to constitute a vital source of both individual and cultural identity and security, a point of departure from which we orientate ourselves in the world."
6
The individual’s sense of place is influenced by his or her place image. As defined by Relph, sense of place is a holistic concept which defies quantitative measurement. Using his methods of analysis, image would also be very difficult to measure. Rclph’s approach is interesting, eloquent and imaginative and has merit. However, it is not intended to have general applicability or to develop "laws’’ of explanation. The approach is descriptive and any attempts at theory development could only be tentative.
In contrast, a number of highly quantitative studies have been developed in the area of choice behaviour and decision-making. In these studies, the images of places have been quantitatively measured in terms of place attributes. Building on the early work of Huff (1960) choice models were diffused through different branches of h'iman geography. These studies were in transport geography (Hcnshcr & Stopher, 1979); retailing (Hudson, 1976; Hanson, 1976; Schuler, 1979) and housing choice (Veldhuiscn, 1988). Some interesting studies were also made in terms of consumers’ spatial preferences. Gould and W hite’s 1974 study on the mapping of mental images and Brown and Moore’s 1970 work on housing preferences are good examples.
Despite the volume of work on image, spatial preferences, and choice, few of the studies discussed so far have linked the concepts of image to behaviour. Pocock and Hudson (1978, p. 3) tentatively suggested that environmental images serve as a means of explaining, possibly predicting behaviour. However, empirical studies of the links between images and behaviour are rare. Such links are usually left as an implicit assumption rather than being explicitly developed (Pocock & Hudson, 1978, p. 13).
1.4 Tourism and Behavioral Studies in Tourist Image and Decision-Making
Just as in geographical research, few studies in tourism research have addressed the linkage between image and decision-making. Mathieson and Wall (1982, p.32) noted this gap in the tourism literature, suggesting that the effects of image of potential destinations on tourist decision-making had received little research attention and deserved more investigation. The same conclusion was reached by Mitchell and Murphy (1991, p. 66) who suggested that "more research needs to be done on the
tourists’ wants, desires and needs" and that "one area that geographers could address is the influence of perception on actual travel choice."
The influence that the tourist image of potential destinations has on resort choice is a new and exciting area of research for tourism geographers since, as yet, we know very little about the process of tourist decision-making. This kind of research is interdisciplinary in nature being also of interest to psychologists, market researchers and tourism managers. It is, however, of central interest to behavioral geographers. As Cadwaliadcr noted, "behavioral geographers are not concerned primarily with the spatial manifestation of behaviour itself but rather with the processes responsible for that behaviour" (1981, p. 6).
As in behavioral geography, research in tourism showed separate strands of development for image research and for decision-making modelling. Image research was largely descriptive, decision making work was more analytical.
A tourist image has been defined as the expression of all objective knowledge, impressions, prejudices and emotional thoughts an individual or group has of a particular object or place (Lawson & Baud-Bovy, 1977, p. 10). This definition is similar to definitions of "sense of place" used in behavioral geography. This is a broad definition and not easy to operationalize. Crompton’s definition of image is perhaps more measurable. "An image may be defined as the sum of beliefs, ideas and impressions that a person has of a destination." (Crompton, 1979, p. 18)
Partly as a response to the problems of measurement, existing research on tourist image has been mainly of the contextual type using content analysis of media advertising (Dilley, 1986; Marsh, 1986; Dearden & Andrcssen, 1987) or postcards (Renwick & Cutter, 1983) or news coverage (Ehemann, 1977). These have attempted to understand the images of tourist places as projected through information sources. However, information sources of this type do not provide a complete picture of destination attributes. For example, the portrayal of the Third World as a paradise today is somewhat inaccurate given the pollution levels and environmental problems in some areas (Britton, 1979). "The tourist experience may be "paradise lost" rather than "paradise found" (Noronha, 1979).
8
In addition, selective perception by tourists easily leads to a stereotyped and inaccurate distorted place image (Moutinho, 1987).
It is the distorted perceptual images which need to be investigated as these form the "behavioral" environment in which decisions are made (Kirk, 1963). Hunt (1975) also emphasized this point. He stated that "it seems likely that images perceived by individuals in the travel market may have as much to do with the area’s tourism development success as the more tangible recreation and tourism resources" (Hunt, 1975, p. 1).
A few image studies have explored this "behavioral" environment, by examining individual preferences and place perceptions (Mayo, 1975; Gartner & Hunt, 1987; Phelps, 1986), but these studies did not relate their findings to behavioral intentions or to actual visits. It remains to be seen whether tourist image distorted by selective perception is a better predictor of behaviour than objective measures of reality. The role of image in influencing tourist product preferences and, ultimately, decisions, is therefore a fruitful area of research (Mathieson & Wall, 1982, p. 2). This study was designed to fill this research gap identified by Mathieson and Wall, by using different methods (as described in Chapter 2) to measure tourist image and to predict destination choice.
The destination choice literature in tourism has been more analytical than the image studies and a number of conceptual frameworks have been developed, mostly at the individual level of decision making. This is because the study of the group decision processes is extremely difficult due to the subtlety of behaviour associated with the interplay of cooperation a».d conflict, persuasion, threat, bluffing and double cross (Green & Wind, 1973).
In choosing a destination a tourist goes though a number of stages in decision-making. The first decision is whether to take a holiday, then what type of holiday to take and finally which resort to choose. This narrowing down of choice to specific locations has been described by Howard (1989), as a hierarchical process where choices are made first at product class level and then at product brand level. A t each stage, decision rules are used to eliminate unsatisfactory alternatives and consumers are left with a limited set o f alternatives which they are considering. These limited sets of destinations
have been referred to as the consumer’s "evoked set" (Howard & Shclh, 1969; Woodside & Sherill, 1977).
The stages in the tourist decision process for the product biand level of choice arc shown in Figure 1.1. First, the tourist forms an image of a place in terms of its natural attractions, amenities and accessibility. Next, some places which arc considered attractive form part of a small "evoked set". These places arc compared and evaluated on their salient attributes and the resort with the greatest utility is hypothesized to be chosen. Choice, therefore, may be viewed as a maximizing utility process, limited by situation constraints, group norms, length of trip planned and available time and financial resources. These situations and social constraints were discussed at length by Moutinho (1987) and Figure 1.2 illustrates and summarises the key points he made in his analysis. Although the individual makes the choices, he or she is strongly influenced by the social environment and situation context in which the decision takes place.
Figure 1.1
Stages in the Tourist Decision Process
TOURIST IMAGE
- the sum of beliefs about a place
- perception of TOURIST PRODUCT in terms of natural attraction, amenities and accessibility
i
EVALUATION OF UTILITY OR OVERALL
ATTRACTIVENESS O F COMPETING TOURIST PRODUCTS IN EVOKED SET
1
Influenced by:
- tourist image, motivation, demographics, lifestyle, distance and benefits sought
1
BEHAVIOURIAL INTENTION
I
DESTINATION CHOICE - a maximizing utility process limited by situation constraints, group norms, length of trip planned, available time, cost
Figure 1.2
Conceptual Model o f Image/Choice Relationships
Situation Context
e.g. length o f trip planned: short break recreation major destination experience
Social Context individual/joint/group decision
T o u r is t_______ s Behavioral Image *--- Intention
family/singles social setting levels of competence in recreation activity
Destination Choice ■
12
1.5 Sample Study
This conceptual framework was used in this research study for a sample of skiers who were residents of Victoria, B.C., Canada and who were in the anticipation stage of planning a skiing holiday (overnight tourist trip) during the 1390 ski season. Their spatial patterns and movements were predicted in relation to their images of the potential resorts held in their evoked sets.
The downhill ski market was chosen as a testing area for investigating consumer decision making for a number of reasons. First, skiing is a clearly defined winter activity taking place at specific locations where facilities are available. Second, it is a large and growing market and one of the most popular outdoor recreation activities in North America. Total skier visits to B.C. resorts from all origins in 1988/89 were 3,629,000 skier days (an increase of 10% on the previous season). Gross revenues front mountain resorts were $116.8 million in 1988/89, which were 39% higher than in 1987/88 season (BC Ministry of Tourism, 1989). The long term trend in skier visits in North America has shown a historical annual growth rate of 8.6% during the period 1977/78 to 1987/88 (Gocldncr, 1989). Therefore, since skiing is such a clearly defined and growing market in B.C., a geographical study involving skier perceptions and destination choice has practical value as Well as academic merit. Such a study will be extremely useful to resorts in terms of product positioning and marketing strategy decisions, and addresses the type of research gap identified by Mitchell and Murphy (1991) in their recent review of tourism geography.
1.6 Research Hypothesis
The specific hypothesis to be tested in this study is that a skier's image of a potential ski resort will influence his or her behavioural intent to visit that resort, which in turn will affect his or her actual choice behaviour. This perceptual-behaviour link is exp. cted to be modified by the length of trip planned (short or long visit) and distance considerations.
1.7 Dissertation Outline
Chapter 2 describes the methodological and theoretical framework for this study. Different approaches to consumer demand analysis arc traced from the normative to probabilistic models, which have developed in a research area which overlaps both geography and marketing. Compositional multi-attribute models and dccompositional conjoint models arc described and their relevance to this research is demonstrated. In addition, a new form of hybrid model is introduced. At the end of Chapter 2 specific research objectives and hypotheses arc summarized.
Chapter 3 describes in detail the methodologies used to address these research objectives. The nature of the research questions dictated a multi-stage research design, with different methodologies for data collection being appropriate at different stages in the research process. Telephone interviews were chosen for prc- and post-trip surveys of the main skier sample. A subgroup of these skiers provided data through a personal interview for the conjoint measurement task. In Chapter 4, the research questions arc used as a framework for data analysis and the presentation of results. Finally, in Chapter 5, a discussion of the research findings is made in terms of their contribution to geographical theory and the practical implications for marketing managers.
CHAPTER 2
THEORETICAL AND METHODOLOGICAL FRAMEWORK
14
The success of research into the spatial significance of perception is necessarily tied to the joint problems of concept identification and measurement. Not only do we have to know which variables arc relevant but we also need to develop appropriate methods for measuring them. (Gollcdge & Rushton, 1976, p. ix)
2.1 Approaches to Model Building
In this chapter the focus is on the academic literature regarding image and decision-making, within the context of paradigm shifts from classical normative models to behavioral probabilistic models, The shift in emphasis from the regional science school’s normative approach to the buyer behaviour school’s individualistic approach in marketing reflects this process (Sheth, Gardner & Garrett, 1988). Within geography, the traditional neoclassical economic approach embodied in "economic man" was, similarly, challenged by the ideas o f procedural rationality suggested by Simon (1976). Pitched at the level of the individual, procedural rationality is concerned with the psychological process of decision-making under conditions of imperfect knowledge. These more realistic behavioral models envisaged man as a "satisficer", making decisions within a limited knowledge and limited choice context (Barnes, 1987, p. 300).
The focus of behavioral models was more on an individual level whereas traditional economic demand models were based on aggregate behaviours and had their conceptual roots in physics. Figure 2.1 shows a framework for classification of the contrasting modelling approaches adopted both in geography and in marketing.
Figure 2.1
A Framework Tor Geography and Marketing Models
School of Thought Aggregate Level Risk Individual Level
Regional Science School
(aggregate models, individual level inferences)
Gravity Models using the laws of social physics Reilly, 1929; Huff, 1963 Risk of ecological fallacy Assumptions: 1. economic man is an "automaton" 2. random processes at the individual level lead to empirical aggregate regularities Consumer Buyer Behaviour School (individual models, aggregate level inferences) Probability estimates to predict benaviour based on grouped individuals or clustered market segments Risk of individualistic fallacy "Satisficcr" concept Cognitive gravity models Multi-attribute models Conjoint models
16 Traditional gravity models have been used to predict interactions between origins and destinations based on the population sizes of the origins and destinations and the distance between them. The basic formula for the model applied to tourism is:
Ty = G P, Aj (1)
i y “
Where: Ty = number of trips made in a given time period between origin i and destination j P, = population of i
Aj = attractiveness of j Dy - distance between i and j
G and a are empirically derived parameters. (Source: Smith, 1989, p. I l l )
One of the major drawbacks of gravity models is that they assume perfect knowledge regarding the attractiveness of competing sites, distances, and costs (Smith, 1983, p. 149). Although they have shown some success in predicting tourist movements at the aggregate level (Stough, 1984), they have revealed nothing about the decision-making process behind the individual behaviour. The gravity model provides useful information about aggregate consumer behaviour for reasons which are not altogether clear at the level o f individual consumers (Bucklin, 1971; cited in Gadwallader, 1981).
In contrast to the traditional gravity model, Cadwallader (1981) suggested a cognitive gravity model approach to predict shopping behaviour. A similar model was used by Perdue (1987) in his study of lake preferences for recreational boating. The latter approach examined the perceived attractiveness of attributes of the tourist product at the individual level. Perdue’s models of lake attractiveness were based on the assumptions of multi-attribute models which have been developed in marketing.
Some attempts have been made to combine demand models at the aggregate and individual levels (Hudson, 1976). However, using results of aggregate models to predict individual behaviour runs the risk o f the "ecological fallacy", in which the characteristics of individuals are wrongly assumed to always have the same characteristics as sample groups within a population. Procedures done vice
versa run the risk of the "individualistic fallacy" (Johnston, 1989, p. 115). Many societies arc more than mere aggregations of individuals and the characteristics of the latter should not necessarily be attributed to the former. Hudson (1976) found that using regression models to predict aggregate choice was only moderately successful in his study of shopping behaviour. As yet, the integration of research predicting aggregate flows from individual choice behaviour has not been adequately addressed. It seems that studies which investigate aggregate tourist flows (e.g. Williams & Zelinsky, 1970) also leave much unexplained, partly as a result of data problems and inadequacies. Williams and Zelinsky, as early as 1970, were postulating that in future research, perception studies should be used to explain tourist flows and that "a question which should be addressed is how potential tourists perceive and evaluate various destinations and how such mental images can actually impinge upon travel decisions" (Williams & Zelinsky, 1970, p. 567).
More recent research has found that the incorporation of behavioral factors into gravity models has increased their predictive capabilities. For example, Styncs, Spoils and Strunk (1985) increased their prediction of park visitor flows by including an awareness variable in their regression model. In this way they achieved a high overall explanation of park flows with an R2 of 0.87 (R2 = coefficient of determination). Perdue (1987) also increased the prediction o f lake visitor flows with the inclusion of simple awareness variables into his predictive models obtaining R2 of 0.94. (This compares to an R 2 of 0.79 obtained when only two independent variables of distance and attractiveness were included in the regression models.) These two studies highlight the importance o f awareness in predicting choice and suggested avenues for further research concerning the integration of models at the individual and aggregate levels.
2.2 Approaches to Mode! Building in This Research
In this research study on skier behaviour the major focus ia on the individual level of explanation. Tourist images of potential ski resorts were measured for individual skiers using a variety o f methodologies which have been developed in marketing. Aggregate probabilities of skier movements were built up from group results of individual skier perceptions and awareness
18 measurements. The theoretical and methodological background for these marketing models is described in the following section. The few tourism studies which have used these methodologies are also reviewed. The approach to research design used in this study follows the steps shown in Figure 2.2.
In the rest of this chapter, a review is made of the different methods by which image may be measured followed by an examination of how the computed image variables may be combined to predict destination choice. Model success is then measured through comparisons with actual skier choice patterns.
2.2.1 Tourist Image Measurement Theoretical Background
The tourist product is a multi-faceted offering, composed of both objective and perceptual attributes. The understanding of a "good", whether a "consumer durable good" or a "service good" as a multi-attribute experience can be traced back to a seminal paper in economics by Lancaster (1966) which was concerned with consumer behaviour theory. Up to that point goods had been undifferentiated and considered as whole products. Lancaster suggested that it was th e characteristics (or attributes) of goods which influenced consumer decision-making, not the goods themselves. The bundle of characteristics were assumed to be more stable than the various products so that predictions could be made successfully on the basis of those characteristics. Consumers were hypothesized to be maximizing utility on the bundle of attributes subject to budget constraints. The ways in which attributes were combined to give a product utility has been investigated with the use o f a number of compensatory, compositional multi-attribute models.
Stages in Research Design
Tourist Image Measurement 1. M ulti-attribute Model 2. Conjoint Measurement 3. Hybrid Model
I
Perceived Attractiveness of Competing Destinations
I
Decision Models to Estimate Probability of Destination Choice 1. Luce’s Choice Model
2. Modified Choice Model
I
20 Multi-attribute Models
Multi-attribute models, based on the early work of Fishbcin (1963) and Rosenberg (1956) were developed in marketing to describe consumers’ attitude towards products. In general, these models have yielded attitude scores which were significantly related to measures of purchase or purchase predisposition (Wilkie & Pessimier, 1973). The basic model is described in the formula (2) below.
n
BI|j •* Ajj ■ B|jk • aik (2)
k = 1
Where: i = consumer (tourist) j = brand (destination)
k = attribute or product characteristic, and n = number of attributes
■ = approximately equals
Bly = consumer i’s behaviourial intention to brand j
Ajj = a unidimensional measure of consumer i’s attitude towards brand j
Bjjjj = the strength of consumer i’s belief that attribute k is possessed by brand j, and
a,k = the degree to which attribute k is desired to exist by consumer i (i.e. its importance)
(Source: adapted from Fishbcin, 1963, p. 233)
This model states that an individual’s attitude towards an object (in this case an image evaluation o f a ski resort) is a function of the amount of valuer’ attributes that an individual perceives the area to have (B|jk) and the importance (aik) of those attributes to the individual. This basic model is based on "cognitive consistency theory" which defines an attitude towards an object as "a composite of the individual’s evaluation o f that object as a means of attaining certain goals weighted by the relative importance or saliency of the goals" (Scott et al., 1978, p. 23).
In later studies, Fishbcin developed his ideas on the multiattributc model into a theory of reasoned action (Ajzen & Kishbein, 1980, p. 5), He recognized that people’s altitudes towards an object may not be strongly or systematically related to their specific behaviours. Rather, the important factor in whether or not consumers will engage in behaviour is their intention to engage in that behaviour. Fishbein modified the basic model to relate consumers’ beliefs and attitudes to their behavioural intentions. According to the thcoiy of reasoned action, people tend to perform behaviours that are popular with other people, i.e. they are influenced by subjective and social norms. A consumer’s subjective norms are measured by the normative beliefs regarding other people’s expectations of an individual’s behaviour and their motivation to comply with the expectations of others (Peter & Olson, 1987, p. 215).
In this study, it was decided to use Fishbcin’s earlier simpler model for analysis mainly because of the complexities involved in measuring the subjective norms influencing behavioural intention. Such complexities would further compound the difficulties of a telephone survey, which by design should be short and simple. In addition, the skier sample was drawn from adults who were already planning a skiing holiday and the focus was not on whether they would ski but rather where they would choose to go.
A number of assumptions are made in the formation o f multi-attribute models. The scales o f measurement for "beliefs" and "importance" have often used rating or ranking scales, and this ordinal data has been treated as if it were interval (i.e. by multiplying and adding scores from rating scales). This assumption is valid according to Labovitz (1970), Labovilz (1972), Cadwallader (1987), and Gaito (1986), on the grounds that although small amounts of error may arise as the difference between two adjacent ranks may not be the same as the difference between two other ranks this disadvantage is offset by the researcher’s ability to use more powerful, robust, and clearly interpretablc statistical techniques. Wilkie and Pessimier (1973, p. 433-434) reported the results of four studies in which different scales of measurement were used and compared with concordance tests. Interval
22 measures (paired comparisons, constant sum scales) tended to yield the same relative ranking of attributes as ordinal measures (rank ordering, rating scales).
Multi-attribute models also assume that attributes are linearly related and arc compensatory. In a linear compensatory model, multiplication of model components procures a linearly related product that is differentially weighted. Low evaluations on one attribute can be compensated for by high evaluations on another attribute. The final image score which is produced by adding the attribute scores is a unidimcnsional measure which represents overall effect (the individual attribute scores arc lost during the summation).
There are only a few studies in the tourism literature in which m ulti-attribute models have been applied and in these studies little attempt has been made to link image with actual behaviour. O ne study by Scott ct al. (1978) did try to link preference for Massachusetts with perceived attractiveness of that state compared with four nearby New England States by using a multi-attribute model framework, but the spatial aspects of variations in tourist image were only poorly defined through a rather crude differentiation of the distance variable into two distance categories. Furthermore, the analysis was related to the attributes o f the state in general rather than of specific resorts. Another study by Goodrich (1977,1978) also used a multi-attribute model to describe the image which American Express card holders had of nine sun destinations. He obtained relative preference rankings of the nine destinations by asking respondents to assign a rank of one to the most preferred destination through to nine for the least preferred. Using Spearman’s rank order correlation coefficient, he found the two measures of destination preference to be significantly correlated (R=0,65, significance 0.05). Again, however, actual choice behaviour was implicitly assumed to be represented by preference and his results apply only to a limited consumer group. In addition, Crompton (1979) used a multi-attribute model to describe spatial patterns in the image of Mexico, but he did not relate his findings to behavioral intentions or to actual visits.
In the present research, a method was developed to use a multi-attribute model as a measure o f tourist image. This method was compared in terms of model structure, predictive ability and
practical application with alternative methods of tourist image measurement (see Figure 2.2). One such method which contrasted with multi-attribute models by taking a dccompositional approach was conjoint modelling. The other method was a hybrid model which combined results from multi attribute and conjoint analysis.
Conjoint Models
Conjoint measurement has been used in choice situations to provide an index of the relative value of attributes or part worths of attributes in influencing preference. Since revealed preference through choice may be the only realistic observable phenomenon accessible to the researcher, this dccompositional approach may have a high degree of validity. Separate "part worths" are estimated for each level of each attribute. These values represent the relative importance of each attribute and its associated level in terms of its contribution to the overall attractiveness of any particular choice (Smith, 1989).
To predict the attractiveness of a product the combination rule for part worths is most often assumed to be additive (Cattin & Wittink, 1982). In this type of model, preference for an object is assumed to be an additive function o', the values (part worths) of its components (attribute levels). This model also assumes a compensatory process for attributes (i.e. respondents will trade off low levels on one attribute for high levels on another).
The conjoint measurement dccompositional preference model for i attributes with each attribute defined at M| levels can be formulated as follows:
i M,
u w - £ E vit •
(3)
i = l k = l
Where: U(x) is an overall utility of preference measure
Vik is the part worth contribution associated with the K'j!' level of the il,,e attribute X|k is the presence or absence of the klhc level of the ilhc attribute
* approximately equals to
24
Since the early 1970s conjoint measurement has been increasingly applied in market research. As yet there have been only a few studies which have used conjoint measurement in tourism and recreation contexts (Cospcr & Kinsley, 19S4; June & Smith, 1987; Woodsidc & Carr, 1988; Filiatrault & Ritchie, 1988). More recent applications of conjoint analysis have assumed interval data and used parametric ordinaiy least squares regression analysis to estimate part worth scores, instead of nonpametric methods c.g. MONANOVA (monotonic analysis of variance). They have also been studied mainly through personal interviews, apart from the Cospcr and Kinsley (1984) study which used a mail survey. However, the Cospcr and Kinsley study achieved questionable results because of limitations in its data collection method.
Some recent studies in conjoint analysis have attempted to reduce data collection volume by experimenting with hybrid conjoint models (Green, 1984; Toy, Roger & Guadagnolo, 1989).
Hybrid Models
In this study, a type of hybrid model was developed, which uses data already collected by multi-attribute and conjoint measurement. It was not developed to reduce data collection but included in the measurement of image for theoretical rather than pragmatic reasons.
The formula for this hybrid model was as follows:
Ay “ X/ B|jk • alk (4)
Such that:
Ay = consumer i’s attitude towards brand j
Bijk = the strength of consumer i’s belief that attribute k is possessed by brand j aik = the degree to which attribute k is desired to exist by consumer i, as revealed by
conjoint analysis
As well as yielding information on the part worth utilities of each level of each attribute, conjoint analysis gives the overall importance score for each attribute in contributing to consumer preferences. This overall importance score when combined with the individual’s strength of belief
score that a destination possesses that particular attribute, gives another measure of imago. This formula is similar to that for the multi-attribute model except that the value of ajk is derived from conjoint measurement.
Previous research has suggested that the prediction of behaviour from attitude variables is improved when situation and context arc clearly specified (Bclk, 1974; Lutz & Bcttman, 1977; Ajzcn & Fishbcin, 1977). In the present research, the three consumer behaviour models were developed within the context of a number of specific factors. The behaviour was skiing and the target object was choosing a ski trip. The situation was the length of the ski trip planned and the time at which the behaviour was to be performed was the next ski trip planned within the 1989/90 ski season.
2.2.2 Decision Rules
The three methods of measuring image in relation to the next ski destination decision were compared and different decision rules were used to predict ski destination choice (sec Figure 2.2). Luce’s choice model (1959) was used as the basic decision rule. Luce’s choice model explains the choice of a destination from a set of possibilities as a function of the perceived utility of that destination in comparison with all other destinations considered. For example, when three destinations are considered the formula is:
P.a = A, (5)
Aa + A b + Ac
P,a = probability of skier in origin i travelling to destination a Aa = perceived attraction of a (defined by image models) Ab = perceived attraction of b
\ perceived attraction of c
This makes sense intuitively although there is no theoretical basis for the rule (Louvi6re, 1988).
26 This basic choice model was modified in two ways to examine the effects of the distance variables of perceived travel time and perceived travel cost. These modified choice models are summarized in the one formula below:
Pi„ = A,
Q a (6)
A + Ab + Ad
D | a Dlb D|C
Where D is a distance variable reflecting the a) perceived travel time or b) perceived travel cost to the destination Pta, A,, Ab, Ac are defined as above, in formula 5.
The modified choice model shown in formula 6, is based on the work of Huff (1963) who developed probabilistic choice models of shopping behaviour. H uffs model subsequently received the attention of a number of researchers. One application in tourism, was the model developed by Wcnncrgren and Nielsen in their study of the recreational use of boating lakes. However, they used "objective" measures of attractiveness and distance rather than perceptual measures as chosen in the present study.
In this dissertation a total of nine models were developed since three choice rules were applied to the image scores from the three image models. Predictions of iki destination choice were then compared with actual choice decisions on both an individual and an aggregate level.
2.3 Research Objectives
Based on the concepts and methodologies reviewed above, the present study is an investigation of the influence of tourist image on destination choice. This research focuses on the downhill skier market segment which originates from Victoria, B.C. The destination awareness of this group, the formation of individual skiers "evoked sets" and the segmentation of skiers according to their length of trip planned are important aspects of this study. The specific objectives of this research are to:
a) choose the salient attributes which could be used to measure ski resort attractiveness and image;
b) use the attributes to develop models for image measurement;
c) use the image scores to produce probability estimates for ski resort visits at an individual level;
d) use the image scores to produce probability estimates to predict aggregate travel movements;
e) compare the image models to determine which model is the best predictor of individual and aggregate travel behaviour;
f) compare the predictive ability of the image models with the basic traditional normative gravity model to determine if the behavioral modification has been an improvement. The approach taken in this research is an exploration of different methodologies derived from both geography and marketing to predict tourist demand. Models from the traditional regional science school are compared with behavioral models which include tourist image measurements. The criteria to be used in evaluating these models will be:
a) how successful they are as predictive devices; and
b) how far they are theory based and deepen our understanding of systems and process. The rationale for this type of research is based on Timmermans’ conclusion that ultimately, the performance of models should be assessed on the basis of comparative empirical studies (Timmermans, 1984, p. 216).
The next chapter describes the methodologies used to address these research objectives as the first step in geographic explanation.
CHAPTER 3
28
RESEARCH DESIGN
3,1 Introduction
This chapter outlines the stages in data collection for a target population of Victoria residents who were in the anticipation stage of planning a skiing holiday during the 1990 ski season.
There were four main phases in the data collection: 1. Prcliminaiy study (January 1990).
2. Pre-trip telephone survey conducted during the ski season (February, March, April 1990). 3. Post-trip telephone survey of respondents from stage 2 (June, July, August, September
1990).
4. Personal interviews for a subgroup of respondents from stage 3 (June, July, August, September 1990).
This approach, illustrated in Figure 3.1 was necessary because of the longitudinal nature of the research and the appropriateness of different methodological tools for different phases in the data collection.
The preliminary study involved a search of the skiing literature to discover what factors were important in ski destination choice and what attributes of ski resorts could be incorporated into preference models to act as good predictors of behaviour. To operationalize the preference models, skiers needed to be contacted while they were in the decision-making phase and data were collected about their beliefs about the resorts in their evoked sets. A telephone survey approach was chosen for reasons of speed, efficiency and cost as the target population was quite difficult to locate (Dillman, 1977). After the ski season, the same sample of skiers was contacted using a quick telephone call to check on their actual ski destination choice behaviour and to make arrangements for individual interviews with a subgroup of the sample. Structured personal interviews were needed to operationalize the conjoint model which involved measuring the importance of resort attributes on destination choice for a hypothetical set of resort profiles. An evaluation of skiers’ beliefs about
Figure 3.1 Research Design
1. Preliminary study to assess the attributes by which skiers rate destinations: data from previous studies and brochures.
3500
List of possible skiers in Victoria (operational population)
Telephone survey randomly sampled from master list. Filter questions
in terms of if they arc planning a ski holiday in Spring 1990
2. Telephone survey questionnaire including multi-attribute model data 3. Telephone survey: actual destination choice data 4. Personal interviews: conjoint measurement data 359 interviews Follow-up telephone survey of original sample 307 interviews 100 interviews
Data analysis and model development
3 0
actual destinations had already been made during the ski season in the pre trip survey. This beliefs data enabled the product positioning by image of actual resorts to be calculated, using the utilities derived from conjoint measurement.
3.2 Preliminary Study: Attribute Generation
"The identification of attributes that consumers will be asked to evaluate is part of the problem definition process, the cornerstone of research design" (Claxton, 1987, p. 466). In this study, it was originally intended to generate the key attributes in ski destination choice using personal interviews and a repertory grid analysis methodology based on personal construct theory (Kelly, 1955). However, a review of the literature on skiing revealed that much work has already been done on attribute generation. Previous work on the ski market was divided into two major areas, a) academic literature, and b) government research.
a) Academic Literature
Previous studies on skiing in which resort attributes were considered, have been largely concerned with the aggregate prediction o f skier demand. Table 3.1 summarizes the results from a number of these early studies in which much use was made of regression models. The number of skiers attracted to competing ski destinations was often used as the dependent variable with a range of independent variables formulated in terms of destination attributes (size, price, access) and slder attributes (demographics, skiing ability). T he predictive validity of these models was variable, ranging from R 2 = 0.29 to 0.80. The early demand studies tended to use variables which were easily quantifiable and they ignored perceptual measurements of image. A notable exception was the Ewing and Kulka (1979) study in which perceived image data were collected for 26 ski resorts in Vermont. Plowever, since data were collected and aggregated from a number of data collection points, the image perceptions are not without spatial bias. By holding the distance variable constant in the analysis and asking skiers to assume resorts were evenly accessible, the authors attempted to address this spatial bias effect. However, they could not address the differing degrees of awareness of resorts resulting
A c a d e m ic S tu d ies o f S k ier D e m a n d S tu d y M e th o d o lo g y D e p e n d e n t V a ria b le In d e p e n d e n t V a riab les Statistical Significance
Size P ric e A ccess S k ier A ttrib u te s
E c h e lb e rg e r & S h a fe r (19 7 0 ) R e g re ssio n ? N u m b e r o f sk ie r days d e m a n d fo r 26 n o rth -e a s t ski re so rts M iles o f in te rm e d ia te trail p e rc e n ta g e o f g ro o m e d slo p e s n u m b e r o f ski in stru c to rs T rav el d ista n c e E is n e r (1971) M u ltip le reg re ssio n se c o n d a ry d ata so u rc e s from U S F S o n site interview s (C a lifo rn ia ) N u m b e r o f trip s to 36 ski re so rts U p hill lift c a p a c ity a m o u n t o f o p e n slo p e T rav el tim e R 2 = 0.40 co n fid en ce level n/a Jo h n so n & E isn e r (1S72) M u ltip le reg re ssio n se c o n d a ry d a ta so u rc e s fro m U S F S o n site interview s (C a lifo rn ia & N ev ad a) N u m b e r o f sk ie r days in 26 ski a re a s
L ife cap ac ity length o f seaso n P ric e c f lift tickets N e a rn e ss to o th e r ski a re a s R 2 = 0.8 c o n fid en ce level 0.05 M c A llister & KJett (19 7 6 ) R e g re ssio n se c o n d a ry d a ta so u rc e s U S F S ski trip d a ta (4 9 ,000 a u to trip s - C a lifo rn ia ) N u m b e r o f ski trip s to 21 m a jo r ski re so rts from 5 8 c o u n trie s L ift carry in g cap ac ity m axim um a ltitu d e m inim um a ltitu d e
T rav el tim e R 2 = 0.69
co n fid en ce level 0.05
Table 3.1 continued A cad e m ic S tu d ies o f S k ie r D e m a n d S tudy M e th o d o lo g y D e p e n d e n t V a ria b le In d e p e n d e n t V ariab les S tatistical S ignificance
Size P ric e A ccess S k ie r A ttrib u te s
E w ing & K ulka (19 7 9 ) R e g re ssio n P re fe re n c e fo r ski re s o r t P erceiv ed length o f slo p p erceiv ed crow ding D ' " " . assu m ed ce m i R 2 = .8 c o n fid en ce level n/a W e tste in & M c N eeley (1980) M u ltip le re g re ssio n n = 784 o n -site interview s (C a lifo rn ia & N e v a d a ) N u m b e r o f ski trip s t o 11 re so rts T rip c o st D ista n c e R 2 = 3 4 co n fid en ce level n/a M o re y (19 8 1 ) L o g it m o d el n = 163 s tu d e n ts in 11 co m m u n itie s N u m b e r o f ski trip s to 15 C o lo ra d o ski a re a s
S ite c h a ra c te ristic s O p p o rtu n ity C o st S kiing A bility m oaified R 2 =
3 7
(goodness o f fit m eth o d ) co n fid en ce level n/a
W alsh & D av itt (19 8 3 ) S te p w ise re g re ssio n n = 8 3 7 o n -site interview s ( A s f e n ) L e n g th o f sta y a t A sp e n C o st p e r day D ista n c e T rav elled D e m o g ra p h ic s R 2 = .29 co n fid en ce level 0.01
