.‘V 'J ' ' O ' ' ' ', ' ‘ ’ _ ‘ ‘ . ' . ‘ 1. I, ’1‘, Illf‘l: ’ Km? .1 ' -' . ", ' 7- -" - - ; ..- ' :fi: him-.1415”.- 1- Zf‘u 5%.. mil f Pug "M ‘ ’WI.V‘ 3' 5;! V .I‘k \‘ffi'. . 7'. w 2‘ “\v. :4 .’fé 3w .‘fi‘l‘f? \j' '1'3: 3“,“ v 'E "In.“ .1." c ’v- 1 .x‘W' M3 F‘L‘ 1"o‘d‘hr) A." .1“ . . . 111.1%». v: 1;": I "In“; g I;'* m fin}? 2,41% 15:.” 91):“? :‘t-gu.‘ fi'fi‘ F... 5.2%“. .5030 ll Ill Illlllllllll Illlllllllllllllllflll 3 1293 10659 7655 , W3” £122? :2? n. . . 9 ‘1 1.1m“ 5"...“ u: ---_"' . ~ . University This is to certify that the thesis entitled THE STABILITY OF THE DIMENSIONS UNDERLYING IMAGES ' OF TOURISM DESTINATIONS IN MICHIGAN presented by David B. Klenosky has been accepted towards fulfillment of the requirements for M.S. degree in Parks and Recreation Resourcesk ; Major péZfessor 5 Joseph Fridgen Date 26 June 1985 0-7639 MS U is an Affirmative Action/Equal Opportunity Institution ' .hdSKJ LIBRARIES .—__ RETURNING MATERIALS: Place in book drdE—to remove this checkout from your record. FINES will be charged if book is returned after the date stamped below. (I? _ M .03 (“13+ :r3 “91" 359 ‘ ii iii. '7 \- ‘4 at‘" ,4 w?- ‘..-— ‘x -I ‘ THE STABILITY OF THE DIMENSIONS UNDERLYING IMAGES OF TOURISM DESTINATIONS IN MICHIGAN By David B. Klenosky A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Park and Recreation Resources 1985 Copyright by David Bruce Klenosky 1985 ABSTRACT THE STABILITY OF THE DIMENSIONS UNDERLYING IMAGES OF TOURISM DESTINATIONS IN MICHIGAN By David B. Klenosky Factor comparison techniques offer a means to investigate the stability of the dimensions underlying visitor images of tourism destinations. This area is useful not only for the development of measurement instru- ments, but for direct marketing applications as well. The purpose of the present study was to test for factor stability: (1) across samples for the same region and (2) across regions for the same sample. Data from the 1982 Frankfort-Tawas Study was tested for factor stability using both visual techniques -- number of factors with eigenvalues greater than one, percent of total variance explained, configuration of factor loadings, factor complexity, and the communalities of the variables -- and vector techniques -- Pearson's correlation coefficient, root mean square (RMS), coefficient of congruence (CC), and the salient variable similarity index. David B. Klenosky Three stable underlying dimensions were identified: Environmental Excitement, Undeveloped Tranquility, and Service Orientation. All had been identified le previous investigations involving destination images. In memory of my beloved grandmother, Rae S. Klenosky. ii ACKNOWLEDGMENTS Since entering my master's program I have become associated with a number of people who have contributed to the completion of this research effort and who have enhanced my personal and professional development. I would like to take this opportunity to express my appreciation to them. First, I would like to acknowledge the support received for this project from the Michigan Sea Grant Program (Project R/R7) and the College of Agriculture and Natural Resources, Agricultural Experiment Station, at Michigan State University (Project 1436-H). Also, I would like to express my gratitude to the faculty and staff of the Department of Park and Recreation Resources for providing the opportunity to continue my academic education in such a dynamic environment. I am grateful to have worked with Dr. Joseph Fridgen, during the past two years. The friendly guidance and support he has provided has made this experience a positive and memorable one. Grateful recognititni is also extended to the other members of my committee, Dr. Dale Wilson and Dr. Daniel Stynes, for their encouragement and advice concerning this investigation and my graduate career. I am indebted to Ed Udd for the information (and careful documentation) he provided regarding the original data collection effort, as well as the assistance he has iii given me throughout my master's program. I'also want to thank James Barron for his patience and assistance during the "too many" hours he put in debugging the data analysis program. Many thanks also go to my parents and family for the confidence they have displayed in me throughout nut educa- tion. Finally, and most importantly, I want to express my sincere appreciation to my fiancee, Laurie Kimble. Her encouragement, understanding, and loving support has truely inspired and sustained me throughout time course of this undertaking. iv TABLE OF CONTENTS PAGE LIST OF TABLES O O O O O O O O O O O O O O O O O 0 O O O Vii LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . x CHAPTER I. INTRODUCTION AND PROBLEM STATEMENT . . . . . . . 1 Introduction . . . . . . . . . . . . . . . . . . 1 Study Objectives . . . . . . . . . . . . . 6 Organization of the Study . . . . . . . . . . . 7 II. LITERATURE REVIEW . . . . . . . . . . . . . . . . 8 Introduction . . . . . . . . . . . . . . . . . . 8 The Uses of Factor Analysis . . . . . . . . . . 9 Image Studies Employing Factor Analysis . . . . 11 Factor Comparison Techniques . . . . . . . . . . 19 Considerations . . . . . . . . . . . . . . . . 20 Objects of Comparison . . . . . . . . . . 20 Substance of Comparison . . . . . . . . . 21 Research Design of the Comparison . . . . 23 Approaches . . . . . . . . . . . . . . . . . . 24 Visual Comparison . . . . . . . . . . . . 24 Vector Comparison . . . . . . . . . . . . 25 Matrix Comparison . . . . . . . . . . . . 30 Studies Employing Factor Comparison Techniques . 32 Research Questions . . . . . . . . . . . . . . . 38 III. RESEARCH METHODOLOGY . . . . . . . . . . . . . . . 39 Research Design . . . . . . . . . . 39 The 1982 Frankfort -Tawas Study . . . . . . . . . 40 Study Areas . . . . . . . . . . . . . . . . . 40 Contact Sites . . . . . . . . . . . . . . . . 42 Sampling . . . . . . . . . . . . . . . . . . . 43 Questionnaire . . . . . . . . . . . . . . . . 44 Study Variables . . . . . . . . . . . . . . . 46 Analysis of Data . . . . . . . . . . . . 49 Reduction of the Variable Set . . . . . . . . 53 Criteria for Comparing Factor Structure . . . . 54 Visual Comparisons . . . . . . . . . . . . . . 54 Vector Comparisons . . . . . . . . . . . . . . 55 IV. RESULTS 0 o o o o o o o o o o o o o o c o o o o o o 57 Comparison Across Samples . . . . . . . . . . . 58 The Subsamples . . . . . . . . . . 58 The Number of Factors Retained for Rotation . 60 Visual Comparisons . . . . . . . . . . . . . . 64 Vector Comparisons . . . . . . . . . . . . . . 83 Comparison Across Regions . . . . . . . . . . . 95 The Sample . . . . 96 The Number of Factors Retained for Rotation . 97 Visual Comparisons . . . . . . . . . . . . . . 101 Vector Comparisons . . . . . . . . . . . . . . 117 Summary . . . . . . . . . . . . . . . . . . . . 128 V. DISCUSSION, CONCLUSIONS, AND RECOMMENDATIONS . . . 134 Discussion and Conclusions . . . . . . . . . . . 134 Study Limitations . . . . . . . . . . . . . . . 142 Recommendations . . . . . . . . . . . . . . . . 144 Summary . . . . . . . . . . . . . . . . . . . . 146 APPENDICES Appendix A. 1982 Frankfort-Tawas Study Questionnaire, Demographic Profile and Summary of Sampling Sites 0 O O O O O O O O O C C I O O O O O I O I O 149 B. Listing of the FACCOMP Data Analysis Program . . . 157 vi 10. 11. 12. 13. 14. LIST OF TABLES Page Appropriate Objects of Comparison for the Possible Research Designs of Comparison . . . . . 24 Breakdown of the Original Sample by Subsample . . . 44 Percent of Respondents Checking Each Adjective as a Descriptor of Either Side "A" or Side "B" by Subsample and Combined Sample . . . . . . . . . . 48 Number of Respondents Completing Each Adjective Checklist (ACL) by Subsample . . . . . . . . . . . 59 Eigenvalues and Percent of Total Variance Explained Before Factor Rotation by Subsample . . 61 Varimax Rotated Factor Matrix for Frankfort Sample of Side "A" . . . . . . . . . . . . . . . . 69 Varimax Rotated Factor Matrix for Tawas Sample Of Side "A" O I O O I O O I I O C O O 1 O O O O O O 70 Varimax Rotated Factor Matrix for Both Samples of Side "A" . . . . . . . . . . . . . . . . . . . 71 Configuration of the Highest Loading Variables for the Frankfort Sample of Side "A" . . . . . . . 72 Highest Loading Variables Ranked by Factor for the Frankfort Sample of Side "A" . . . . . . . . . 73 Configuration of the Highest Loading Variables for the Tawas Sample of Side "A" . . . . . . . . . 74 Highest Loading Variables Ranked by Factor for the Tawas Sample of Side "A" . . . . . . . . . . . 75 Communalities by Subsample . . . . . . . . . . . . . 81 Variables Classified into Communality Thirds (Upper, Middle, and Lower) by Subsample . . . . . 82 vii Table Page 15. Factor Comparison Across Samples Using the Correlation (Pearson's r) of Factor Loadings . . . 84 16. Factor Comparison Across Samples Using the Root Mean Square (RMS) of Factor Loadings . . . . 86 17. Factor Comparison Across Samples Using the Coefficient of Congruence (CC) of Factor Loadings . . . . . . . . . . . . . . . . . . . . . 88 18. Factor Comparison Across Samples Using the Salient Variable Similarity Index (S-Index) of Factor Loadings . . . . . . . . . . . . . . . . 91 19. Factor Comparison Across Samples Using the Correlation of Factor Loadings, RMS, CC, and S—Index o o o o o o o o o o o o o o o o o o o o o 93 20. Number of Adjective Checklists (ACL' 5) Completed by Subsample . . . . . . . . . . . . . 96 21. Eigenvalues and Percent of Total Variance Explained Before Factor Rotation by Region . . . . 98 22. Varimax Rotated Factor Matrix for the Combined Sample of Side "A" . . . . . . . . . . . . . . . . 103 23. Varimax Rotated Factor Matrix for the Combined Sample of Side "B" . . . . . . . . . . . . . . . . 104 24. Varimax Rotated Factor Matrix for the Combined Sample of Both Side "A" and Side "B" . . . . . . . 105 25. Configuration of the Highest Loading Variables for the Combined Sample of Side "A" . . . . . . . 107 26. Highest Loading Variables Ranked by Factor for the Combined Sample of Side ”A . . . . . . . 108 27. Configuration of the Highest Loading Variables for the Combined Sample of Side "B" . . . . . . . 109 28. Highest Loading Variables Ranked by Factor for the Combined Sample of Side "B". . . . . . . . . . 110 29. Communalities by Region . . . . . . . . . . . . . . 115 30. Variables Classified into Communality Thirds (Upper, Middle, and Lower) by Region . . . . . . 116 31. Factor Comparison Across Regions Using the Correlation (Pearson's r) of Factor Loadings . . . 118 viii Table 32. 33. 34. 35. 36. 37. 38. 39. .40. BI. Factor Comparison Across Regions Using the Correlation (Pearson's r) of Factor Scores Factor Comparison Across Regions Using the Root Mean Square (RMS) of Factor Loadings Factor Comparison Across Regions Using the Coefficient of Congruence (CC) of Factor Loadings . . . . . . . . . . . . . . Factor Comparison Across Regions Using the Salient Variable Similarity Index (S-Index) of Factor Loadings . . Factor Comparison Across Regions Using the Correlation of Factor Loadings, Correlation of Factor Scores, RMS, CC, and S-Index Highlights of the Results for the Comparison Across Samples and the Comparison Across Regions . . . . . . . . . . . . . . . . Configuration of the Highest Loading Variables for Dimension 1 7- Environmental Excitement Configuration of the Highest Loading Variables for Dimension 2 -- Undeveloped Tranquility Configuration of the Highest Loading Variables for Dimension 3 -- Service Orientation Demographic Profile by Subgroup ix Page 120 121 123 124 126 132 156 LIST OF FIGURES Figure Page 1. Scree Test for Frankfort Sample of Side "A" . . . . 63 2. Scree Test for Tawas Sample of Side "A" . . . . . . 63 3. Scree Test for Combined Sample of Side "A" . . . . . 99 4. Scree Test for Combined Sample of Side "B" . . . . . 99 CHAPTER I INTRODUCTION AND PROBLEM STATEMENT Introduction The importance of tourism to Michigan may be measured by the extent of the state's investment in tourism promotion. The Michigan Travel Bureau's "Yes Michigan!" promotional campaign (formerly "Say Yes to Michigan") is one indicator of that investment. At present, the state's investment in tourism promotion alone totals approximately nine million dollars (Morris, 1984). 'The "Yes Michigan!" campaign has been designed to attract tourists to vacation in Michigan. Attracting tourists to the state offers a means for: (1) stimulating employment, (2) increasing tax revenues, (3) (Liversifying the state's economy by attracting new industry, and (4) improving public perceptions, both within the state and nationally, of Michigan as a good place to live, visit, and do business (Morris, 1983). The last item, improving the image and perceptions of the state, is central to the campaign to encourage travel in Michigan. Monitoring travelers' images and perceptions of the state is necessary to determine if the "Yes Michigan!" campaign has been effective in achieving that goal. 1 2 The images held by potential visitors about an area may have a significant influence upon the viability of that area as a tourism destination (Hunt, 1975). It is hypothe- sized that regardless of whether an individual has ever visited a particular region, various forces have had an impact on how that individual perceives the region. Paid advertising, news stories, special interest stories, conversations with friends and relatives, as well as one's past experiences all combine to shape one's image of a given area. The image that forms is likely to be as significant as more tangible regional amenities when vacation decisions are made. Information concerning destination images can be used to: improve’a negative image, build upon a positive image, or correct a distorted one. Although several studies have explored regional images, the criteria considered relevant for the examination of a destination's image do not appear to be concrete entities. Different studies have utilized a variety of dimensions to operationalize the construct image. Hunt (1975) considered respondent impressions of a destination's: residents, climate, and attractions. In two other studies (Gearing, Swart and Var, 1974; Var, Beck and Loftus, 1977) five criteria were utilized: natural factors, social factors, historical factors, recreational and shopping facilities, and, lastly, infrastructure and food and shelter. Pearce (1982a) used 21 set of constructs descrip- tive of a destination's: scenic beauty, tourist offerings, 1m 3 social and political climate, seasonal sport offerings, and suitability for different vacationer "styles" (e.g. seclu— sion versus high-contact, relaxation versus adventure, etc.). While the dimensions used to study image appear to be quite diverse, there seems to be a degree of commonality among them as well. A better understanding is needed of the components underlying a destination's image to determine if these components are stable across different regions and across different groups of people. An objective investigation of the components underlying the images of tourism destinations is important both for methodological and marketing purposes. From a methodological viewpoint, the criteria used in) assess the image of an area must first be determined in order to develop operational image measurement instruments. The need to develop reliable and valid measurement instruments in tourism studies has received attention from Pearce (1982b), and, in Michigan, from industry representatives in the state (Fridgen, 1982). As Goodrich (1978) suggests, such an instrument can be used to monitor changes, if any, in perceptions of tourism destinations over tin”: through longitudinal studies, or, as mentioned above, to compare the perceptions of selected tourist areas held by various groups of tourists (e.g. visitors versus potential visitors), or to compare perceptions of different regions (e.g. the Lake Michigan coast versus the Lake Huron coast) by the same group. 91 4 From a: strategic marketing perspective, the dimen- sions underlying regional images may be used in at least two ways. The first is in product positioning. As stated by Kotler (1984), product positioning is the act of designing the organization's product and marketing mix to fit a given place in the consumer's mind. Thus, the identified dimen- sions of image can be used by the Michigan Travel Bureau to position Michigan against the offerings of competing Great Lakes states, cur, on another level, position destinations within the state to appeal to a variety of tourist segments. The second marketing-related application is in the formulation of advertising messages and strategies.' Identifying the underlying criteria used to assess a destination's image would be an obvious benefit to efforts aimed at developing promotional strategies for those destinations. The relative importance of those criteria can be examined, and the more salient ones can then be stressed in advertising messages aimed at the proper market segment. In short, for both methodological and marketing reasons, the study of the dimensions underlying regional images is an important research area. Although the study of regional images is important, it is at the same time complex. Therefore, it is necessary to limit the scape of the proposed research. The regional diversity within Michigan complicates the study of destination images. The proposed study will be exploring basic regional images, but will be developing and 5 comparing dimensions based on coastal regions onlyq 'The question that will.run: be addressed is whether the image structure of non-coastal regions is different than the image structure of coastal regions. While such an examination would be beneficial, it is not among the objectives of this research. The proposed study, then, will focus on images of coastal regions only. 9 Two regions in Michigan's coastal zone will provide the data for this analysis -- the northeast (Lake Huron) and northwest (Lake Michigan) coasts of Michigan's lower peninsula. Interest in these areas developed during a study conducted by Michigan State University during the summer of 1982 (Fridgen, Udd and Deale, 1983). People indicated a real preference for the northwest coast and virtually ignored the northeast coast when completing a cognitive mapping task asking them to circle areas they felt best provided for recreation and tourism opportunities in the state. The 1982 Frankfort-Tawas Study, was subsequently undertaken to investigate whether there were differences in the perceptions and images of the two coasts. Respondents at each location (Frankfort and Tawas) were each asked to rate two different regions of Michigan's Lower Peninsula (the northern Lake Michigan coast and the northern Lake Huron coast) using an adjective checklist format. The use of this format offers the potential to explore the dimen- sions underlying regional images with the use of factor analytic techniques. Further, the research design employed 6 in the study allows the examination of regional images across groups of respondents and across target regions. In summary, the focus of this research is on examining the nature and stability of the dimensions underlying regional images of specific coastal destinations in the state of Michigan. A twofold approach will be used to achieve this goal. In the first step, tine Frankfort- Tawas Study, discussed above, will be factor analyzed to identify the dimensions which underlie visitor images of the two coastal areas in question. The second step will test the stability of those dimensions across different groups of respondents (those in Frankfort and those in Tawas) for the same region (e.g. the Lake Michigan coast), and across different regions (the Lake Michigan coast and the Lake Huron coast) for the same group (e.g. the Frankfort sample). Study Objectives The general objective of this study is tn) explore the stability of the dimensions underlying regional images. Relative to this general objective, the specific study objectives are: 1. Identify the underlying dimensions of the image of two coastal regions in Michigan. 2. Test the stability of those dimensions across different respondents and different regions. 7 Organization of the Study The remainder of this study is organized into five chapters. A review of the relevant literature is the subject of the next chapter. The literature review includes sections on: the uses of factor analysis (in general and in regional image studies), the techniques used to compare factor structure, and applications of those techniques to studies in tourism and recreation. The chapter concludes by presenting the research questions used to guide the data analysis portion of the study. Chapter III presents the research methodology. This includes a summary of the original data collection effort on which the present study is based, the study variables, the procedures used to analyze the data, and the criteria which will be used to evaluate the results of the factor comparisons. Chapter IV offers the results of the factor comparisons. This chapter consists of two major sections: comparison across samples and comparison across regions. The fifth and final chapter contains a discussion of the results, conclusions, study limitations, and recommendations for future research in this area 0 CHAPTER II LITERATURE REVIEW Two objectives for this research are outlined in the previous section. The first objective is to identify the underlying dimensions of the image of two coastal regions in Michigan. Factor analysis will be used for this phase of the analysis. The second objective tests the dimensions identified 1J1 the previous objective for stability across different groups of respondents and across different regions. This phase will employ factor comparison techniques. To more fully understand the methods applied in this research, this section will provide discussions of: (1) the uses of factor analysis; (2) studies that have utilized factor analysis to investigate regional images; (3) the techniques used to test for the stability of factor struc- ture; and (4) tourism and leisure research studies employing factor comparison techniques. o—«l re fa 9 The Uses of Factor Analysis An implicit assumption of factor analysis discussed by Stewart (1981) is of special interest to this study. He notes that a chief assumption of factor analysis is that "major differences found in everyday human relationships become part of the language of the culture. At one level factor analysis is concerned with how people use the language, its words, concepts, etc. and the empirical relationships within the language. The underlying assump- tion is that these empirical relationships within the language will reveal something about human behavior on another level" (p. 51). Thus, the theoretical basis of factor analysis is that people use language in similar ways. The techniques of factor analysis are used tn) operationa— lize this theoretical concept. "The term factor analysis refers to a broad category of approaches to conceptualizing groupings (or clusters) of variables and.eu1 even broader collection of mathematical procedures for determining which variables belong to which groups" (Nunnally, 1978, p. 327). As stated by Harmon (1976), the chief aim of factor analysis is "to attain scientific parsimony or economy of description" (p. 4). Thus, factor analysis is most often thought of as a data reduction technique. As an data reduction technique, applications of factor analysis can be classified into one of the following categories: (1) exploratory uses -- the exploration of and 10 detection of patternings of variables with121 view to the discovery of new concepts and a possible reduction of data; (2) confirmatory uses -- the testing of hypotheses about the. structuring of variables in terms of the expected number of significant factors and factor loadings; and (3) uses as a measuring device -- the construction of indices to be used as new variables in later analysis (Nie, Hull, Jenkins, Steinbrenner, and Bent, 1975). In the present study, factor analysis will be used mainly for exploratory purposes —— to identify the underlying dimensions of the images of two coastal regions in Michigan and to determine the extent of in”: stability of regional images across groups of respon- dents and across regions. It is important to recognize that factor analysis does not indicate quantitative differences between varia- bles, although these may indeed be important. Rather, it is used to explore the dimensional structure for data; it indicates the important qualities present in the data. In short, factor analysis provides a means for reducing the number of variables in a study without great loss of information and serves to identify the important qualitative distinctions in the data (Stewart, 1981). 11 Image Studies EmployingiFactor Analyst; Studies which employ factor analysis to assess regional images are, for the most part, sparse in the literature. It appears that the technique has been utilized either only as a first step in an analysis or used to examine only a single component of a region's image. Nevertheless, it is instruc- tive to present these applications of factor analysis. Crompton (1979) utilized factor analysis to develop a set of 30 semantic differential statements to investigate Mexico's image as a tourism destination. Crompton's goal was "to develop a comprehensive set of terms which, when taken together, would constitute a valid content universe of the image of Mexico" (p. 19). The set of terms was derived in a two step proce- dure. The first step entailed a content analysis of selected general reading materials on Mexico and of adver- tising brochures published by the Mexican National Tourist Council. Eight content areas of area image emerged and key words or phrases descriptive of image attributes were collated. Crompton then expanded this set of words and phrases using a series of 36 unstructured interviews which were similarly content analyzed; In the second phase of his procedure, Crompton orga- nized the basic descriptive terms into 42 semantic differen- tial statements which were administered to a convenience sample of students (n - 70). After factor analyzing this data, Crompton reduced the 42 statements to a set of 30 by 12 retaining those which displayed 21 salient loading (”1 the factors. The statements were subsequently administered to a sample of students across the U.S. (n = 617) to determine the relative importance of each image attribute and to explore the relationship between respondents' geographic location upon those attributes. This study could be criticized because neither the eight content areas of image nor the procedures employed in the factor analysis were noted explicitly. Further, it was not stated whether the eight content areas of image identi- fied in the first step were validated in the factor analysis used in the second; one might expect this to be the case. Finally, the number of cases used to reduce the number of semantic differential statements in step two (N a: 70), was too small given the number of statements/variables (42) which were factor analyzed. The "rule of thumb" for factor analysis is five cases for each variable (Kass and Tinsley, 1979, p. 124). Craik (1975) employed an adjective checklist format to understand and account for the individual differences in landscape descriptions rendered by participants who took an auto tour through an "everyday physical environment located in Marin County, California" (p. 131). The sample used in this study was considered reasonably representative of the general population of Marin County (n s 187). The research project entailed a multi-step proce- dure. Individuals were first given the auto tour through 13 the target site. Following the tour, they were asked to describe the place they toured using a 240 item environmen- tal adjective checklist. In addition, they were asked tn) describe themselves on a number of standard personality and attitude measures. The analysis of the data for this study can be broken down into two phases. The first centered on the adjective checklist (ACL). Of the 204 items on the ACL, only those items receiving endorsement rates greater than 10 percent were retained, yielding a subset of 104 items. These adjectives were then submitted to a principal axis factor analysis, employing the "highest r" method of estimating communality (this method places on the diagonal for a given row of the correlation matrix the largest correlation coefficient in that row). Varimax rotation of a four factor solution accounting for 65.2 percent of the variation in the data resulted in four factors. The four factors were labeled (1) serene-gentle; (2) dry-barren; (3) beautiful-picturesque; and (4) blooming-cultivated. Factor scores were then computed, and used in the next phase. Phase two used the factor scores for tflua four descriptive landscape dimensions identified in phase one as the inputs for a typological analysis. This analysis used a hierarchical clustering algorithm (BCTRY OTYPE) to develop clusters of individuals. Craik was able to identify 16 types based on the landscape descriptions. As a final step, he proceeded to report, in textual format, the characteris- Ct to 14 tics (as reported on the personality and attitude measures) of the respondents comprising each group. This study offers a great deal of guidance both methodological and theoretical. Methodologically, Craik employed factor analyses on an adjective checklirnz to develop summary descriptions of landscapes. This method parallels the proposed research. In addition, the retention of adjectives with endorsement rates greater than 10 percent has received support from other empirical studies factor analyzing dichotomous variables (Chase and Cheek, 1979). Although dichotomous variables are not strictly consistent with the assumptions of factor analysis, using dichotomies with splits less than 10 percent indicates an extremely skewed distribution. Such a distribution is obviously very far from a normal one, which factor analysis assumes is the case. Theoretically, Craik was able to show that different groups of respondents can be grouped into typologies based on differences in how they perceive the same environment. In this case, the researchers assumed that factor structure was stable across respondents. The question-the proposed research raises and seeks to investigate is whether this assumption is a valid one. This issue is one of the chief goals of the proposed analysis. Ritchie and Zins (1978) explored the importance of culture as it relates to the attractiveness of Quebec as a tourism region. The respondents in the study were managers 15 and functionaries from various sectors of the tourism system in Quebec. 'The study relied on a set of eight "general" factors, which were originally developed through an expert judgment approach by Gearing, Swart, and Var (1974) to determine the overall attractiveness of tourism regions. The eight factors were: (1) natural beauty and climate; (2) culture and social characteristics; (3) sport, recreation, and educational facilities; (4) shopping and commercial facilities; (5) infrastructure of the region; (6) price levels; (7) attitudes towards tourists; and (8) accessibi— lity of the region. A factor analysis was performed, using interval scale measurements, on the set of eight "general" dimensions yielding eight independent factors. Though some correlation was reported between the dimensions accessibility and infrastructure as well as between the dimensions price levels and commercial facilities, the researchers concluded that all eight dimensions should be used separately in evaluating the overall attractiveness of a tourism region. Within the cultural and social dimension, twelve elements were hypothesized to contribute to the attractive— ness of a tourism region -- (1) handicrafts; (2) language; (3) traditions; (4) gastronomy; (5) art/music; (6) history; (7) work; (8) architecture; (9) religion; (10) education; (11) dress; and (12) leisure activities. Respondents were asked to rate the above elements as "perceived" from the standpoint of both residents and non-residents. 16 Separate factor analyses were run on the 12 varia- bles for ratings received from both tflue resident and non—resident perspective. Only the factors derived from the resident perspective were interpreted and presented. This analysis yielded four dimensions which explained 64% of the variation in the set of cultural variables. These were labeled: (1) elements of daily life; (2) remnants of the past; (3) the good life; and (4) work. Unfortunately, neither the type of factor analysis employed nor the method of rotation (if any) were mentioned. This information would have been useful in the present study. Pizam, Neumann, and Reichel (1978) sought to empirically identify the dimensions underlying tourist satisfaction with a tourism area and suggest methods to measure them. 'This study was chiefly concerned with feelings of gratification or displeasure about a destination following interaction with that destination. Although regional image studies, including the proposed research, normally explore the additional aspect of how regions are perceived by individuals regardless of whether they actually visited them, this research is very instructive. The sample in this study consisted of summer tourists to Cape Cod, Massachusetts (n - 685). The first step in the analysis was to develop a means to measure the construct of tourist satisfaction (con- ceptually defined as a collection of tourists' attitudes about specific domains in the vacationing experience). 17 Seven major domains were identified (based on 51 review of the literature, consultation with experts familiar with Cape Cod and the study of tourism, and a series of Open-ended interviews with tourists): (1) accommodations; (2) eating and drinking establishments; (3) accessibility; (4) attrac- tions; (5) cost; (6) amenities and facilities; and (7) hospitality. Thirty-two 5-point Likert scales were develop- ed to operationalize the range of satisfactions embodied in those domains. Of the 32 Likert scales items, those receiving the highest ratings were natural assets -- scenery, natural attractions, the environmental quality, and beaches -- and tourism facilities including hotels, motels, auui restau- rants. The items receiving the lowest ratings included high costs, traffic conditions, and extent of commercializa- tion. I Factor analysis was used to identify the dimensions of tourist satisfaction with destination areas. The authors write: "Factor analysis is especially useful in measuring tourist satisfaction since the tourism product is made up of many interre- lated components each of which requires a separate measure of satisfaction. By using the factor analytic technique we can simplify the multiplicity of these measures" (p. 317). In this study principal factoring without iteration (principal components analysis) was performed on the 32 items. Factors with eigenvalues (before rotation) greater 18 than or equal to one were retained, and varimax rotation was specified. The procedure above yielded 8 factors. Twenty—four of the original 32 items had loadings of .60 or greater with one of the eight factors. The 8 factors were labeled (1) beach opportunities; (2) cost; (3) hospitality; (4) eating and drinking facilities; (5) accommodation facilities; (6) campground facilities; (7) environment; and (8) extent of commercialization. Evidently, there is a good deal of overlap between the identified factors and the hypothesized satisfaction domains. The authors note that their findings by no means suggest that the above factors are universal. They propose that the factors probably depend on a number of elements including the destination area, its facilities, attractions, land formations, weather, and so forth. They conclude, however, that destinations having features similar to those of Cape Cod, Massachusetts -- rural summer beach resort areas -- could use the same factors as the ones developed in this study. Given, the importance of the Great Lakes coastal zone and the similar seasonality of tourism in Michigan, theirs' is a notable conclusion. All the studies discussed above have employed factor analysis either as an initial step for further analysis or to summarize regional images into a set of underlying dimensions. None of these studies have sought to compare factor structure across groups of respondents or across 19 regions. Alternately, studies which have employed factor comparison techniques have not been concerned with regional images (to date). The techniques employed in factor comparison studies is the subject of the next major section. Factor Comparison Techniques The comparison of the results of separate factor analyses entails an examination of the consistency of those results. Such comparisons have inherent linkages to the development of a science as a body of knowledge. As Rummel (1970) notes: 'Hha build a science requires that findings be sufficiently explicit to make possible evaluation, replication, and comparison with other studies. Each study in its own right may contribute a bit of knowledge -- a datum -- to building a science. But these data output of different studies must be integra- ted into general propositions and given meaning in terms of a theoretical framework. This requires that comparison between find- ings be possible so that the replicable substantive patterns can be identified, and the unique, research-design-specific results can be discarded" (Rummel, 1970, p. 449). The factors that result from a particular analysis cannot be considered definitive until some form of analysis is performed to assess their generalizability. "A factor once found remains merely a hypothesis about a pattern; it is verified only after the pattern has been found again in well-defined circumstances" (Cattel and Baggaley, 1960, p. 33). Factor comparison techniques provide a means for determining the stability of factors across samples, ’(J ’1, /\ r! (A 20 variables, and experiments. The purpose of this section is to discuss the techniques of factor comparison suggested by the literature. In all methods of factor matching the variety of preliminary decisions made 1J1 the course CHE the analysis should be the same across the two or more studies being compared (Rummel, 1970). This entails consistency in: the scaling procedures used on the original data, associational statistics (correlation measures), and factor extraction procedures. The idea is to ensure that the degree of dissimilarity between the studies can be attributed to differences in interdependencies, not differences in methods. "As a general rule, comparability of procedures removes one obvious plausible rival source of variance" (Levine, 1977, p. 43). Considerations Before an appropriate approach can be determined a number of elements must first be considered in contemplating or making a comparison. These include: the objects of comparison, the substance of comparison, and tflua research design underlying the comparison. Object of Comparison. Five possible matrices may be produced from a factor analysis (Rummel, 1970): (1) correla— tion inverse matrix, (2) factor loading matrices, (3) higher-order matrices, (4) factor regression matrices, and (5) factor score matrices. Although each one may be suitable for a particular purpose, only two will be consid- 21 ered here: factor loading matrices and factor score matri- ces. Factor loading matrices give the loadings of the variables on factors. There are two major types of matrices, tflua initial or unrotated factor matrix and the simple structure or rotated factor matrix. The unrotated matrix defines the patterns of variance in the data; the rotated matrix identifies the clusters of intercorrelation among the variables. Among the rotated matrices, comparison may be of either orthogonal or oblique factor matrices. Further, for the oblique case one must consider the choice between pattern and structure matrices. If the cases as well as variables are of interest, factor scores also may be compared between studies. Use of this technique is suggested when the subjects are the same since "similarity of factor loadings across two structures does not imply similarities of factor scores across the two structures necessarily" (Rummel, 1970, p. 457). Substance of Comparison. Comparison may involve several aspects of the factor results. While one aspect, such as the loadings, may predominate in the comparison, a full assessment of the similarity of two studies should involve other considerations as well (Rummel, 1970). This entails comparisons of: factor configurations, complexity, variance, number of factors, and communalities. The configuration of variables refers to the pattern and magnitude of the factor loadings. Rummel (1970) 22 points out that configuration comparisons are the most common form of factor comparison. Complexity is another aspect that may be expressly compared. Rummel (1970) provides some guidance for this area of comparison: "Although implicitly involved in the comparison of configurations, tine relative complexity of variables and factors may be explicitly noted. Does a variable highly loaded in one study spread across factors in another? Does a specific factor in one study shift to a common factor in another? Shifts in complexity are clues to the underlying differences in the data. They provoke "why" questions, answers to which may explain some of the differences between the studies" (p. 453). One may also wish to compare the variance accounted for by each and all of the factors. When the factors are consistently found to explain considerable variance across studies, this suggests that the factors represent some underlying dimension rather than being an artifact of a particular study (Allen and Buchanan, 1982). In addition, the number of significant factors extracted from different studies may be compared to provide an indication of the degree of convergence on the underlying dimensions. Moreover, such comparisons can be helpful in assessing the best number of factors to rotate in future analyses. Finally, an analysis of the communality of a variable between studies may be conducted. This approach allows one in: determine which variables are highly inter- correlated and which are consistently unique. 'The latter 23 may result from poor measurement characteristics, poor data, or causal influences not under the control of the investiga- tion. The identification of these consistently unique variables encourages such questions and may provoke research to answer them (Rummel, 1970). Research Design of the Comparison. The type of comparison contemplated is largely dependent upon the nature of the research design. The possible situations of compari- son are: (1) the variables and cases CHE both studies are the same; (2) the variables are the same, but the cases differ; (3) cases are the same, but the variables differ; run! (4) neither variables nor cases are the same (Rummel, 1970). In the first instance above, factor loadings and/or factor scores may be used as the object of comparison. These will determine the extent the relationships between variables are the same between studies. For the second situation, in which several variables are similar but cases differ, comparison is restricted to the factor loading matrices. When several cases are the same, but the varia- bles differ, as in the third situation, only factor scores may be compared (this relates back to the previously mentioned point that similarity of factor loadings does not necessarily imply similarity of factor scores across two structures). Finally, with different subjects and different variables "the problem belongs to Alice in Wonderland" (Cattel, Balcar, Horn, and Nesselroade, 1969, p. 782). That 24 is, unless some linkages are established apart from the factor analysis, this type of factor comparison is inappro— priate. Table 1 presents a summary of the appropriate objects of comparison for the possible research designs of the comparison. Table 1 Appropriate Objects of Comparison for the Possible Research Designs of Comparison CASES VARIABLES Same Different Same - Factor Loadings - Factor Loadings - Factor Scores Different - Factor Scores Approaches The approaches used to compare factor structure can be classified into three groupings: (1) visual comparisons; (2) vector comparisons; and (3) matrix comparisons. These approaches increase in complexity moving from the first to the third. Visual Comparisons. This approach to factor comparison entails the visual matching and assessment of the results from various studies. It involves comparing the configuration, communalities, and complexities of the variables and factors to get an overall impression of their 25 agreement. This observational analysis allows one to look for subtle differences between factors, which can then be augmented by more mathematical techniques of comparison (Rummel, 1970). Another type of visual approach suggested by Levine (1977) is to pool the two matrices adding a dummy variable indicating group membership (that is, perform one factor analysis on the combined sample using an additional varaible which identifies the subsample each respondent is from). "The loading of this identification variable would indicate those factors, assumed to exist in both groups, on which the groups are most discriminated, therefore, the factors for which the group's mean factor scores would be most different. This technique does not give separate factor structures for the two (or possibly more) groups but one should always assume that the cases from two or more groups are homogene- ous until this assumption breaks down empirically (p. 43). This technique is not applicable to situations where one does not access to the original data -- for example, if one is comparing one's own data to another previously published study, the data from which is not available. Vector Comparison. The mathematical approaches available for comparison can be divided into those that compare several pairs of factors and those that compare the whole factor matrix. This latter approach will be discussed in the following section. The vector comparison approach takes the factors in the different studies as they are.. No attempt is made to compensate for their peculiar errors, specific variances, and effects of dissimilar variables 26 (Levine, 1977). Thus, factors are compared in pairs and their similarity measured by one of the techniques discus- sed: The most commonly employed technique used to compare the loadings (or factor scores) between factors, when the same variables are involved, is the product moment correla— tion coefficient. This approach, however, does suffer from a number' .1) HY - hyperplane (-.l < fi < .1) NS a negative salient (fi < -.1) cij - the number of loadings in cellij The index may be seen as a comparison of the difference between the number of hits and misses as a proportion of a weighted sum of the cell frequencies. It reaches a maximum value of +1.0 (when there is a perfect relationship) and a 30 minimum of -l.0 (when there is 21 perfect inverse relation- ship); a value of 0 indicates no similarity between the two factors. Thus, the values obtained are easily interpreted (like the product moment correlation coefficient). Of course, the index is throwing out a large amount of informa- tion, that is, it equates a loading of .9 with one of .3, but it also reduces the risk of capitalizing on chance differences among the loadings. Probability tables are available which provide critical values for significance testing. These tables are organized according to the number of variables under investigation and by the percentage of loadings between the two factors falling into the hyperplane category. "Since this index makes sense, is so easy to calculate, and has an approximate test of significance, I suggest it strongly as one of (hopefully) several measures 'used" (Levine, 1977, p. 48). Finally, there is another approach that has received attention in published empirical studies. 'This approach, which appears appropriate when comparing 2 sets of factors derived from two samples on the same set of variables, is performing an analysis of variance on the factor scores derived from the separate analyses. While this method appears in the empirical literature, no reference to it appears in the more technical factor comparison literature. The approach, however, does seem to be a valid one. Matrix Comparison. The final approach to factor comparison determines the extent of the match between two 31 factor matrices. This approach addresses the problems associated vdJfli vector comparisons. "The difficulty with the vector comparison approach is that the factors are compared as given. Exogenous influences may affect the independent rotations of the two studies and confound tflua comparisons" (Rummel, 1970, p. 463). There are two main methods of matrix comparison, canonical analysis and target analysis (also known as transformation analysis). A discussion of these techniques, however, is beyond the scope of this research. Of the myriad of factor comparison techniques available, one suggestion is consistent throughout the literature: employ several, do not rely on one alone. "By employing a combination of approaches, the limitations CHE individual techniques are minimized and the potential for nusinterpreting factor similarities is substantially reduced" (Allen and Buchanan, 1982, p. 310). Prior to discussing the specific methodology to be employed in the proposed study, it is useful to examine the procedures that have been applied in previous Studies using factor compari- son techniques. 32 Studies Employing Factor Comparison Techniques Studies employing factor comparative techniques emerged primarily from the disciplines of psychology and political science. Similar studies in the leisure and tourism fields do exist and are instructive for this research. 'This sectirni'will present those studies which' offer some guidance on the techniques used to compare factor structure. In recreation and leisure research, the emphasis of these studies has been either on testing the stability of leisure motivations across samples or assessing the con— gruence between participation and preference/interest for various leisure activities. The discussion of these studies will mainly stress their technical aspects. In tourism only one investigation has been identified which is related to factor comparison. Although it should be noted that this study does not utilize any of the comparison techniques described above, it is pertinent to the discussion of comparative research. Stringer (1984) described the aspects of six graduate thesis including the work of McCullough (1977), who focused on images of tropical destinations. McCullough used the concept image in an empirical investigation of the similarities and differences in the images of tropical "holiday" destinations held by experienced "long-haul" travelers and the images attributed to those destinations by travel agents. The travelers (n u 56) were asked to imagine 33 a perfect holiday on an unnamed tropical island and to use a Q-sort method on a set of SO descriptive statements to determine the relative salience to them of each statement. Similarly, travel agents (n = 21) were asked to sort the same items according to their assessment of the relative importance of each item to a typical, experienced client. Two factor analyses were performed, one (”1 the traveler sample alone and one on the entire sample. The first produced 11 image dimensions which were labeled: (1) romantic-practical; (2) service oriented-cultural; (3) peaceful-festive; (4) functional-comfortable; (5) reliable- adventurous; (6) public-private; (7) secure-foreign; (8) social-physical; (9) natural-sophisticated; (10) convenient- exotic; and (11) organized-unspoilt. McCullough inferred that all these dimensions contrasted people's normal, civilized life-style, involving notions of familiarity, efficiency, security, and privacy, with the uninhibited, natural, or uncivilized life-style of a remote holiday -— specifically, romance, festivity, and the exotic. The second factor analysis performed was a Q-mode analysis (using the observations as variables). It produced two groups, one mainly of travel agents the other 'of travelers. The differences between the two sets of images were found to be mainly on items located in the factors romantic-practical, and natural-sophisticated. The travel agents overemphasized the importance to tourists of sun- shine, sophistication, and romance, and underestimated 34 the importance of unspoilt nature, local culture, and social encounters. While this study was not consistent with any of the factor comparison techniques presented in the previous section, it does demonstrate two points. First, research in tourism offers many unique opportunities to employ factor comparison techniques in meaningful ways -- McCullough emphasized that travel agents could learn unufli from the differences in tflua relative importance attributed to the various images. Second, variations between groups of respondents were found, in terms of how they evaluate tourism environments. As mentioned before, such an investi- gation is central to the proposed research. Graefe, Ditton, Roggenbuck, and Schreyer (1981) sought to examine the dimensionality, stability, and the importance of motives for participating in the same recrea- tional activity -- river floating -- in two different environmental settings'---- the Green and Yampa Rivers in Dinosaur National Monument on the Colorado- Utah border (n - 854) and on the Rio Grande River in Big Bend National Park in Texas (n - 253). The researchers used a mailed questionnaire distri- buted to a sample of river users who had obtained a (manda— tory) river use permit. Data was collected on a 38 item, multidimensional motive scale (developed by Driver) using a six-point response format to indicate the relative impor— 35 tance of each motive as 21 reason for going (Hi the river trip. In this study, factor analysis was used to address the issue of factor stability across samples. The method of principal factoring with iterations (principal axis), followed by varimax rotation of factors with eigenvalues greater than one, was used to reduce the 38 items to a smaller number of orthogonally unique motive scales. The coefficient of congruence (CC) -— a vector comparison approach -- was used to test the similarity between the 7' factors derived from the Green and Yampa sample and the 8 factors from the Rio Grande sample. In short, certain motive constructs were found to be more stable across the samples than other motive constructs; these were characterized as learning/experiencing nature and stress release/solitude. Other constructs were not as stable, nor was the stability of the original hypothesized constructs verified. The study can be criticized in that it only used one technique -- the coefficient of congruence -- to assess the similarity of factor structure across samples. Bishop (1970) factor analyzed the results of a survey concerned with the frequency of participation (on a nine-point scale) in 25 recreation activities in four selected midwest communities -- Minneapolis, Minnesota (n a 925) and in Illinois, River Forest (n s 130), Glencoe (n = 411), and Elk Grove (n a 415). 36 The data for each community was subjected to a factor analysis (principal axis) using squared multiple correlatdxui coefficients as the communality estimates. A three factor solution (determined by those factors with eigenvalues greater than one) was then rotated according to the varimax criterion. Estimated factor scores were then calculated for all respondents. The stability of factor structure was assessed using a visual comparison approach. As a final step, Bishop correlated the factor scores for each subject with demo- graphic and socioeconomic variables to lend support to the factor interpretations. Bishop was able to demonstrate that the three factors -- active-diversionary, potency, and status -- were stable across the four communities. This study did not employ any technical factor comparison techniques to assess stability: An analysis of the data using those techniques would help in) verify the results found. Finally, Allen and Buchanan (1982) illustrated the use of five factor comparison techniques -- the correlation of factor scores, the correlation of factor loadings, the S-index, the root mean square (RMS), and the coefficient of congruence (CC) -- by comparing a leisure factor structure based on participation data with a leisure factor structure based on interest dated ‘The authors measured respondents' interest and participation in 52 specified leisure activi- ties (chosen to represent 17 leisure categories identified 37 through previous factor analytic studies) using five point Likert scales. Separate factor analyses were performed for both groups of variables (interest and participation) using principal factoring with iterations (principal axis). The initial common factors were rotated to a final solution using varimax rotation. Only factors with eigenvalues greater than one were retained for discussion euui further analysis. The resulting nine factors for the interest variables were then compared to the eight factors from the participation data using each of the five techniques above -— all vector comparison approaches. The results indicated that six of the factors were very similar for the two factor structures. They concluded that by using the five measures above, one achieves a more sensitive analysis of factor structure similarity than can be achieved by standard correlational measures alone. This study was instructive in that it employed almost all of the factor comparison techniques discussed in the previous section. Summary This chapter was primarily concerned with acquain- ting the reader with the techniques used for comparing factor structure and the applications of those techniques to tourism, leisure, and recreation issues. The chapter. included sections on the uses of factor analysis, applica- tions of factor analysis to studies of regional images, the 38 techniques used to compare factor structure across studies, and, finally, the applications of those techniques in leisure and recreation. Research Questions The present study is exploratory. In recreation, factor comparison research is a relatively new area; in tourism, it is even newer. Therefore, specific hypotheses will not be tested explicitly. Instead, the following research questions were developed to guide the data analy- sis: (1) Are any of the dimensions underlying tourists' images of a specific tourism destination stable across different samples of respondents?, and (2) Are any of the dimensions underlying tourists' images of tourism destinations stable across different destinations (target regions) for the same sample of respondents? CHAPTER III RESEARCH METHODOLOGY Research Desigp The literature review suggested a number of techni- ques tn) compare factor structure in a variety of research designs. It demonstrated the use (Hf these methods in: several leisure and recreation studies, and emphasized that the application of these methods to assess tflua stability of regional images of tourism destinations has not been previously attempted. To reiterate, this study is directed at one basic question: what is the stability of the dimensions which underlie regional images across different subsamples (Hf respondents and across different target regions. To address this question, the ideal study would employ a research design that samples (at least two) different subgroups' images of (at least two) different regions using an established measurement scale specifically designed to assess regional images. Mainly due to the time and monetary constraints of collecting primary data, secondary data was used in the present research. Specifi- cally, a pilot study -- the 1982 Frankfort-Tawas Study -u- 39 40 from a larger tourism image project intended to identify and assess traveler-defined tourism regions in Michigan (see Fridgen, Udd, and Klenosky, 1984; and Fridgen and Klenosky, 1985) provided an appropriate data set for the analysis. Although not a perfect substitute, the 1982 Frankfort—Tawas Study does meet the research design requirements and, for the most part, the measurement requirements as well. The 1982 Frankfort-Tawas Study This section summarizes the 1982 Frankfort-Tawas Study which provided an appropriate data base for the present study. The Frankfort-Tawas Study was designed tx> explore the images and perceptions of two distinct coastal regions (the Lake Michigan and Lake Huron coastlines of the northern lower peninsula) held by visitors to two distinct destinations (Frankfort, 16cated on the Lake Michigan coast, and Tawas, located on the Lake Huron coast). This study was a pilot study intended to provide baseline information for use in subsequent studies in.zi broad tourism image project conducted by the Department of Park and Recreation Resources at Michigan State University. The original objective of the study was to compare images and perceptions of two selected areas in Michigan and to make inferences from any similari— ties or differences that ocurred. Study Areas For the proposed research, which entails a compari- son of samples drawn from Frankfort and Tawas, it is 41. important to understand why the tun) communities were originally selected as study areasl. First, each community has a relatively small population (1,967 for Tawas and 1,603 for Frankfortz) which indicates that the two areas are of approximately equal size. The second reason for selecting these two communi- ties as study sites is that they provide and service a large variety of waterbased and non-waterbased recreational opportunities. The Frankfort area is located a: few miles south of Sleeping Bear Dunes National Lakeshore which serves as a major recreational resource in Michigan providing opportunities for: duneclimbs and hikes, scenic drives, beaches and swimming, canoeing, fishing, euui camping. In addition, two major rivers flow into Lake Michigan in this area -- the Betsie River to the south and the Platte River to the north. Several inland lakes with public access are also located in this area -- Crystal Lake, Platte Lake, and Little Platte Lake -- which provide for additional water- based activities. Finally, the Huron Manistee National Forest provides major tracts of forested lands for public recreational use. The Tawas area possesses a similar set of natural attractions. Lake Huron provides for all forms of boating and other waterbased activities. Tawas Point State Park is 1The majority of this discussion is taken from Eckstein (1983). 2Michigan 1980 Census of Population. 42 located at the tip of Tawas Bay and provides beaches, swimming, hiking, and camping facilities. Also located in the Tawas area are: three campgrounds (with over 400 sites); two public fishing docks; charter fishing boats; boat launching sites; and riding, hiking, snowmobile, and cross-country ski trails. Inland waterareas in the Tawas area include Tawas Lake, the Tawas River, and tflua AuSable River. Finally, the Huron National Forest, located along the AuSable, provides for a variety of recreational opportu— nities. Both shoreline areas possess an abundance of natural resources providing recreation opportunities on a Great Lake or in surrounding inland areas. In addition, the two communities are of a comparable and manageable size. And finally, the two areas were used as study sites in a previous tourism study with favorable results (see Eckstein, 1983; and Eckstein and McDonough, 1983). Contact Sites Four sites in the Frankfort area and five sites in the Tawas City/East Tawas area were selected with the help of Sea Grant Agents in those areas. Contact sites were selected based on two primary factors: (1) to find potential questionnaire participants who were visiting the study area (either Frankfort or Tawas) on a vacation trip; and (2) to provide an adequate number of participants for sampling. In short, the sites selected were intended to provide a representative sample of tourists/vacationers in the two 43 study areas (see Appendix A for a summary of the sampling sites utilized in the study). Sampling The survey population consisted of individuals, age twelve and older, who were vacationing in the Frankfort and Tawas areas during August, 1982. Data was gathered on four consecutive days from August 16—19, 1982. The questionnaire was self administered. Potential respondents, age twelve and older, were approached by members of the research team and asked to complete the survey. Where groups of potential respondents were encoun- tered, one person in the group was randomly selected to complete the survey. Individuals agreeing tn) partici— pate were given a survey on a clipboard. If respondents had difficulties with the survey, members of the research team were nearby, available to provide assistance. The sampling method varied slightly depending on the site being sampled. When there was a site where most or all of the people had to pass by, the data was collected by distributing surveys to people passing by. When people at a site were more sedentary, the data was collected by having the interviewer move across an area once or twice to pass out surveys. The resulting sample was essentially a census of an area or of all the people passing by a specific point. This census was conducted under the limitation of one person only being able to manipulate six clipboards at one time thus 44 allowing some people to get up and leave or pass by without receiving £1 questionnaire. .A total of 287 questionnaires were collected over the four day period with at least twenty questionnaires from each of the nine sites. Table 2 presents a breakdown of the original sample by subsample. Table 2 Breakdown of Original Sample by Subsample SUBSAMPLE n* Frankfort 136 ( 47.4) Tawas 151 ( 52.6) Total 287 (100.0) "Numbers in parentheses indicate the percentage. Questionnaire The questionnaire used in this study contained two major sections. The first section, was designed to opera- tionalize visitors' images and perceptions of the two regions. The secbnd section provided descriptive and socioeconomic information. The two major sections were comprised of the following types of questions: 45 SECTION I. REGIONAL IMAGE/PERCEPTION INFORMATION - Adjective Checklist - Side "A" - Adjective Checklist - Side "B" - Attribute Checklist/Comparison SECTION II. DESCRIPTIVE/SOCIOECONOMIC INFORMATION - Rating of Side "A", Side "B", and Michigan 1- Familiarity with Side "A" and Side "B" - Trip Purpose - Residence Status — Socioeconomic Information (Sex, Age, Education, Income) Since the emphasis in the study was on respondents' images and perceptions of Michigan, the initial set of instructions on the survey informed respondents that "we are interested in what you thigh about Michigan. It is not important that you have not been to all of the regions in the state to answer the questions". The first task asked of respondents was to complete two adjective checklists —— one for each coastal region. A forced choice checklist of regional attributes was then presented which asked respon— dents to indicate which of the two sides ("A" or "B") best provided for each of the set of attributes listed. The remainder of the survey asked respondents to provide descriptive and socioeconomic information. A complete discussion of the original survey and sampling procedures can be found in Udd (1982). Also, for a further discussion of the study consult Fridgen and Klenosky (1985) (a copy of the original questionnaire can be found in Appendix A). 46 Study Variables The variables of interest in the present study,are the adjectives that make up the two adjective checklists (ACL's). The two adjective checklists were identical except that one focused on the northwest, or Lake Michigan, coast. of the lower peninsula (to be referred to as Side "A" in the remainder of this analysis) and the other focused on the northeast, or Lake Huron, coast of the lower peninsula (to be referred to as Side "B"). The adjective checklists were created using three inputs: an adjective checklist used in.ei previous environ— mental perception study by Craik (1975), brochures of tourist attractions in Michigan, and discussions with the Sea Grant agent in Tawas. The adjectives fell roughly into two categories: descriptions of the area and descriptions of the social situation in the area. Antonyms for these words were used to create descriptive pairs. One adjective from each pair was randomly selected to be included in the survey. Lastly, the adjectives were randomly assigned to a location (”1 the survey. Tb offset any ordering bias, on half of the questionnaires distributed, respondents were‘ requested to describe the northeast coast first; on the other half, respondents were requested to describe the northwest coast first. The instructions for the adjective checklists read as follows: "The following is a list of adjectives. Please read them quickly and check each one you would consider 47 descriptive of coastal area "A" (or "B") shown on time map at left". The adjectives were coded as dichotomous data. A one was recorded if the adjective was checked by the respondent and a two if the adjective was not checked. The adjectives, in the order originally presented in the survey, are displayed in Table 3. This table shows the percentage of respondents, for each subsample and for the combined sample, checking each adjective as a descriptor of each side of the state. “- - n an... '0 Qt: 8pc IBI In In: col ill! “I? 111 lpp 48 Table 3 Percent of Respondents Checking the Adjective as a Descriptor of Either Side "A" or Side "B" By Sub-Sample and Combined Sample -—--COMBINED ----- ‘ ---—FRANKFORT ----------- TAWAS ------ ADJECTIVE "A" SIDE "B" SIDE "A" SIDE "B" SIDE "A" SIDE "B" SIDE secessible 73.52 36.02 49.02 74.82 60.62 56.42 clean 83.82 31.62 60.32 77.52 71.42 55.72 crowded 9.62 8.82 20.52 23.22 15.32 16.42 secluded 28.72 14.02 13.92 12.62 20.92 13.22 drab 0.02 5.92 0.72 2.02 0.32 3.82 flat 1.52 11.82 2.62 10.62 2.12 11.12 expensive 24.32 10.32 34.42 13.92 29.62 12.22 open 27.92 12.52 18.52 23.82 23.02 18.52 forested 72.82 24.32 34.42 42.42 52.62 33.82 friendly 72.82 25.72 44.42 70.92 57.82 49.52 unusual 15.42 2.92 7.92 3.32 11.52 3.12 peaceful 78.72 25.72 44.42 62.32 60.62 44.92 pleasant 78.72 26.52 51.72 68.22 64.52 48.42 siddle class oriented 36.82 20.62 16.62 45.02 26.12 33.42 sandy 79.42 14.02 42.42 62.92 59.92 39.72 ugly 0 02 0.72 0.72 0.72 0.32 0.72 sonotonous 0.02 2.92 0.72 1.32 0.32 2.12 clear 39.72 8.12 18.52 32.52 28.62 20.92 hostile 0.02 0.72 0.02 1.32 0.02 1.02 fun 73.52 20.62 55.02 64.92 63.82 43.92 fasily oriented 75.72 28.72 46.42 72.22 60.32 51.62 spirited 16.22 4.42 18.52 25.22 17.42 15.32 accepting 27.22 8.82 15.92 26.52 21.32 18.12 courteous 47.12 16.92 25.22 42.42 35.52 30.32 gracious 25.02 9.62 16.62 20.52 20.62 15.32 enjoyable 75.72 27.92 53.02 72.82 63.82 51.62 tacky 2.22 1.52 1.32 2.62 1.72 2.12 spectacular 31.62 7.42 17.92 9.92 24.42 8.72 prisitive 17.62 1.52 7.32 7.32 12.22 4.52 resote 14.72 6.62 7.92 6.62 11.12 6.62 scenic 87.52 32.42 59.62 68.92 72.82 51.62 unspoiled 40.42 10.32 18.52 24.52 28.92 17.82 colorful 61.82 19.12 49.72 52.32 55.42 36.62 quaint 19.92 2.22 7.92 11.92 13.62 7.32 upper class oriented 17.62 2.92 21.92 4.62 19.92 3.82 alive 30.12 7.42 18.52 25.82 24.02 17.12 appealing 60.32 20.62 38.42 47.02 48.82 34.52 bright 33.82 8.82 18.52 25.22 25.82 17.42 cossercial oriented 12.52 5.12 19.92 11.32 16.42 8.42 delightful 52.22 10.32 26.52 36.42 38.72 24.02 exciting 33.12 8.12 21.22 30.52 26.82 19.92 festive 15.42 6.62 15.22 15.22 15.32 11.12 outdoor oriented 77.22 24.32 43.72 60.92 59.62 43.62 horrible 0.02 0.02 0.02 0.72 0.02 0.32 out-of-the-way 11.02 5.92 11.92 6.62 11.52 6.32 lifeless 1.52 2.22 0.02 0.72 0.72 1.42 nondescript 0.72 4.42 0.72 1.32 0.72 2.82 noisy 2.22 1.52 2.62 6.62 2.42 4.22 interesting 55.12 15.42 37.72 39.12 46.02 27.92 tourist oriented 53.72 18.42 50.32 53.02 51.92 36.62 quiet 47.82 14.02 19.22 31.82 32.82 23.32 natural 66.22 22.12 38.42 47.02 51.62 35.52 restful 64.72 19.12 35.12 47.02 49.12 33.82 developed 11.02 10.32 18.52 16.62 15.02 13.62 tasteless 1.52 1.52 2.02 2.62 1.72 2.12 classy 10.32 0.72 5.32 4.02 7.72 2.42 heavy traffic 8.82 8.12 14.62 15.92 11.82 12.22 busy 12.52 5.12 20.52 22.52 16.72 14.32 N . 136 N - 151 N - 287 49 Analysis of Data The investigator adhered to the following proce— dures.in the data analysis stage of this research: A. The number of variables to be analyzed in the adjective checklist was first reduced to those receiving endorsement rates of 102 or greater from all subsamples. This procedure follows the work of Craik (1975) and, in addition, helps improve the performance of the phi coefficients (discussed below). Four separate factor analyses were performed: 1) Tawas of Lake Michigan, (2) Frankfort of Lake Michigan, (3) all respondents of Lake Huron, and (4) all respondents of Lake Michigan). The first two factor analyses, (1) and (2) addressed the first research question -- factor stability across samples; while analyses (3) and (4) focused on the second research question -- factor stability across regions. Since the data embodied in the adjective checklist is dichoto- mous, rather than using standard pearson product moment correlations to form the initial correla- tion matrix, phi coefficients were used. The use of phi coefficients rather than other correlational measures has received attention from Chase and Cheek (1979), Kim and Mueller (1978), and Rummel (1970) (Preliminary runs 50 using the product moment correlation and phi coefficient formulas yielded identical results -- the same coefficients -- in fact, for dichotomous variables both use the same formu- las. Therefore, though the theoretical suita- bility of phi coefficients is noted, this issue is not central to the analysis). Principal axis factoring was used as the method of initial factoring. This technique is similar to principal components except that communality estimates are used in the main diagonal of the correlation matrix rather than ones. Further, the use of correlation coefficients computed from dichotomies does not violate tine assump- tions of the principal axis factor model (Gorsuch, 1974). Squared multiple correlation coefficients were used as the initial communal- ity estimates. The number of factors to be retained and rotated was determined by examining the number of factors with eigenvalues greater than one, the percent of variance explained by each factor, scree tests, and interpretability. The initial factor solution was rotated according to the Varimax criterion to aid iJl obtaining simple structure. The factor analyses were performed with the use of the Statistical Package for the Social Sciences 51 (Nie, Hull, et al., 1975) on the CDC Cyber 750 computer at Michigan State University. Factors were not labeled (interpreted) until the tests for the stability of factor structure (outlined below) were completed. 9 Criteria were established for assessing the extent of the match between structures prior to testing for factor structure stability. Factor structure was first be tested for stabi- lity across samples for the same target region (Tawas of Lake Michigan and Frankfort of Lake Michigan) by comparing the factor loadings matrices. Use of the factor loadings matrix is appropriate when comparing factor structures from different sets of cases on the same set of variables. The following comparison approaches were used; visual comparisons (configura- tion, complexity, variance explained, number of factors, and communalities) and vector compari- sons (Pearson's product moment correlation, root mean square, coefficient of congruence, and the S-index). Factor structure was then tested for stability across target regions (Lake Huron versus Lake Michigan) for the same sample (in this case the. combined sample). For the most part, all the comparisons in this section used the factor 52 loadings matrices. In addition, however, the correlation of factor scores was also computed. Comparison of factor scores is called for when comparing the factor structure of the same set of cases across two sets of variables. Problems with generating factor scores from dichotomous data (Kim and Mueller, 1978) are recognized, and thus, may be a limitation of this analysis. The following comparison techniques will be used: visual comparisons (configuration, complexity, variance explained, number of factors, and communalities) and vector comparisons (Pearson's product moment correlation -- of the factor loadings and factor scores, root mean square, and coefficient of congruence). A Pascal program was written to facilitate the calcula- tion of the root mean square, the coefficient of congruence, and the S-Index for both the comparison across samples and across regions. A listing of that program (FACCOMP) is included in Appendix B. Once the factors which remained stable across samples and regions were identified, they were interpreted and named. 53 Reduction of the Variable Set As discussed in the previous chapter, the first step in the analysis was to reduce the original set of adjectives to those which received attention by at least ten percent of the respondents. Gorsuch (1974) suggests that dichotomous variables with splits (the percentage of 1's versus the per- centage of 0's) beyond 102/902 or 902/102 not be included in factor analysis because they could too severely limit the potential range of the phi correlation coefficient. Previous studies factor analyzing dichotomous data have also followed this procedure (Chase and Cheek, 1979; Chase, et al., 1980; and Craik, 1975). In the present study, the majority of the total sample of tourists attributed the following characteristics as descriptors of the two regions: scenic, accessible, clean, pleasant, enjoyable, family-oriented, fun, peaceful, and outdoor-oriented. The array of additional noteworthy descriptors is considerable. A total of 27 of the 58 adjectives received endorsements by at least ten percent of the total sample of visitors to the two areas (Table 3). (None of the original adjectives were endorsed by more than ninety percent of the total sample.) These 27 adjectives are intended to operationalize respondents' images and perceptions of the two coastal areas -- they comprise the data set used for the remainder of the analysis. 54 accessible middle-class-oriented appealing clean sandy delightful secluded fun outdoor—oriented expensive family oriented interesting open courteous tourist—oriented- forested enjoyable quiet friendly scenic natural peaceful unspoiled restful pleasant colorful developed Criteria for Comparinngactor Structure A major question that arises is how does one interpret the results of the various techniques used to assess factor similarity? Specifically, what criteria does one look for to identify invariant factors? As Chase, Kasulis, and Lusch (1980) point out, there is little practical guidance in the literature. For this research, which employs a number of techniques to compare factor similarity, no single compari- son measure will be used to determine the stability of a given factor. Instead, the results of all the methods of comparison will be considered before assessing a: factor's stability. Nevertheless, to guide the analysis, it is necessary to establish operational criteria for each similarity measure. Relevant to establishing this criteria, the visual approaches to factor comparison will be presented first, followed by the vector approaches. Visual Comparisons The visual methods of comparing factor structure are: the'number of factors (with eigenvalues greater than one), configuration of factor loadings, factor complexity, 55 percent of variance explained, and communalities of the variables. For these methods the criteria for factor stability are somewhat subjective. As long no substantial deviation exists between the factor matrices on each of these factor comparison measures, factor structure cmnx be considered stable. Specific criteria for each measure are as follows: 1. Number of factors with eigenvalues greater than one -- +/—1 between matrices. 2. Configuration of loadings -— at least two variables loading highly on a given set of factors between matrices. 3. Complexity -- for the variables -- the same complexity of a variable on a set of factors across matrices (i.e. the same variable loading highly on two factors on two matrices); for the factors -— a factor found in one matrix is identifiable in one or more factors on the second matrix. 4. Percent of variance explained -- +/- 2 percent between factors and +/— 5 percent between matrices. 5. Communality of the variables -— variables whose communality remains in the same third across two matrices. (For this comparison the variables will be divided into three communality categor— ies -- top third, middle third, and bottom third -- depending on the magnitude of the communality for a given variable). Vector Comparisons As outlined in the literature review, the vector or factor to factor comparison approaches are: Pearson's product moment correlation coefficient, root mean square (RMS), coefficient of congruence (CC), and the salient variable index (S-Index). These methods entail a more 56 objective means of comparison than the visual comparisons euui require specific criteria to assess factor similarity across factor matrices: 1. Pearson's correlation coefficient (r) -- correlation coefficients greater than .40 that are statistically significant at the Prob < .05 level. Root mean square (RMS) -- coefficients between 0 and .10. Coefficient of congruence (CC) —— coefficients in absolute value of .80 or greater. Salient Variable Index (S-Index) -- coefficients in absolute value of .80 or greater. CHAPTER IV RESULTS This chapter is divided into two main sections. The first section addresses research question (1); the compari- son of factor structures across samples. In the second section, the focus is on research question (2), comparing factor structure across regions. Each section employs both visual and vector factor comparison approaches to assess the stability of tflue factor structure underlying respondents' images of tourism destinations. 57 58 Comparison Across Samples To address the first research question, this section focuses (n1 determining whether any of the criteria which underlie respondents' images of tourism destinations remain stable across different samples of respondents. The research design in this type of factor comparison involves the same variables and different groups of respondents. The variables used in this analysis were the reduced set of 27 adjectives discussed in the previous chapter. The formula- tion of the two subsamples of respondents used for this comparison is the topic of the following section. The remainder of this section is divided into three major headings: the number of factors retained for rotation, visual comparison approaches, and finally vector comparison approaches. The Subsamples The first step at this stage was to develop the sub- samples of respondents for input into the factor analyses. Two criteria were considered in creating the subsamples. The first criteria was homogeneity within each subsample. The respondents comprising each subsample had to be as alike as possible. The second criteria was sample size. Each subsample had to be large enough to assure that the results of the factor analyses were reliable. Homogeneity. Respondents were first assigned to one of two groups depending upon their survey site -- Frankfort or Tawas. To insure homogeneity within these two samples, 59 only respondents who checked at least one adjective in a checklist were retained. In this way, only respondents who were familiar enough with the target region were input into the factor analysis. The number of respondents who complet— ed an adjective checklist (ACL) for each coastal region by subsample is presented in Table 4 below. Table 4 Number of Respondents Completing Each Adjective Checklist (ACL) by Subsample* ACL FOR ACL FOR SUBSAMPLE SIDE "A" SIDE "B" Frankfort 134 ( 51.7) 67 ( 31.5) Tawas 125 ( 48.3) 146 ( 68.5) Total 259 (100.0) 213 (100.0) *Numbers in parentheses indicate the percentage. Sample size. The second criteria used to create the subsamples was size. The size of each subsample had to be large enough to ensure that the correlations used as the input for the factor analyses accurately reflected the correlations for the underlying population of tourists. The . generally accepted "rule of thumb" in factor analysis is to have at least five subjects for each variable being measur- ed, with an absolute minimum of 100 subjects (Kass and Tinsley, 1979, p. 124). 60 Therefore, to factor analyze the 27 regional image variables the sample size had to be roughly 135 (27'varia- bles times 5 cases per variable). Obviously, the number of respondents in the Frankfort sample completing 2n1 ACL for Side "B" (N u 67) is well below the suggested minimum of 100 cases, precluding a comparison between samples for this side of the state. However, the sample sizes for the two groups completing an ACL for Side "A" is over 100 (for the Frankfort subsample N = 134 and for the Tawas subsample N = 125) and close enough to the benchmark of 135 to serve as the two subsamples for this particular comparison. In summary, the variables used for the comparison of factor structure across samples are the 27 adjectives used to describe Side "A" -- the northern Lake Michigan coast- line. The two subsamples consist of 134 respondents surveyed in Frankfort and 125 respondents surveyed in Tawas. The next section discusses the decision regarding the number of factors to retain for factor rotation and subsequent factor comparison. The Number of Factors Retained For Rotati22_ The 27 adjectives above were then factor analyzed to identify the dimensions underlying respondents' images of the target region (Side "A") for each subsample. The two factor analyses indicated that the patterns of interrela- tionships in the data were very similar (Table 5). The decision to make at this point centered on the number of factors to extract for rotation for both subsamples. Four 61 methods to determine the number of factors to retain were considered: the number of factors with eigenvalues greater than one, percent of variance explained by each factor, scree tests, and interpretability. Table 5 Eigenvalues and Percent of Total Variance Explained* Before Factor Rotation by Subsample PERCENT OF CUM. PERCENT OF TOTAL VARIANCE TOTAL VARIANCE EIGENVALUES EXPLAINED EXPLAINED FACTOR FRANKFORT TAWAS FRANKFORT TAWAS FRANKFORT TAWAS 1 6.564 6.422 24.32 23.82 24.32 23.82 2 1.914 1.885 7.12 7.02 31.42 30.82 3 1.600 1 689 5.92 6.32 37.32 37.02 4 1.431 1.587 5.32 5.92 42.62 42.92 5 1.252 1.392 4.62 5.22 47.32 48.12 6 1.197 1.196 4.42 4.42 51.72 52.52 ' 7 1.117 1.175 4.12 4.42 55.82 56.82 8 1.018 1.055 3.82 3.92 59.62 60.72 9 0.924 1.014 3.42 3.82 63.02 64.52 10 0.890 0.915 3.32 3.42 66.32 67.92 11 0.835 0.842 3.12 3.12 69.42 71.02 12 0.829 0.785 3.12. 2.92 72.52 73.92 13 0.740 0.740 2.72 2.72 75.22 76.72 14 0.695 0.692 2.62 2.62 77.82 79.22 15 0.675 0.639 2.52 2.42 80.32 81.62 16 0.632 0.614 2.32 2.32 82.62 83.92 17 0.617 0.592 2.32 2.22 84.92 86.12 18 0.578 0.560 2.12 2.12 87.12 88.12 19 0.525 0.478 1.92 1.82 89.02 89.92 20 0.475 0.457 1.82 1.72 90.82 91.62 21 0.443 0.444 1.62 1.62 92.42 93.22 22 0.423 0.409 1.62 1.52 94.02 94.82 23 0.387 0.394 1.42 1.52 95.42 96.22 24 0.377 0.297 1.42 1.12 96.82 97.32 25 0.324 0.274 1.22 1.02 98.02 98.32 26 0.291 0.234 1.12 0.92 99.12 99.22 27 0.244 0.218 0.92 0.82 100.02 100.02 * Based on principal axis factoring. 62 The Number of Factors with Eigenvalues Greater Than 033. The "rule of thumb" in factor analysis is to extract and rotate only those factors that have eigenvalues greater than one. If one uses this criterion, however, the Tawas sample should have nine factors retained and the Frankfort sample eight (Table 5). For the Tawas sample, at nine factors retained, 64.52 of the total variation in the data is explained; for the Frankfort sample, at eight factors, 59.62 is explained. Percent of Variance Explained. Another "rough" criterion specifies that each factor to be rotated explain at least five percent of the total variation in the data. Using this cutoff, five factors should be extracted for the Tawas sample and four for the Frankfort sample (Table 5). Scree Test. Another method for determining the number of. factors to extract for rotation is to perform a scree test (Cattell, 1966). A scree test is basically a plot of the eigenvalues on the factors. One looks for the point of discontinuity on the plot, the point where the "scree" begins, to determine the cutoff for the number of factors to retain for rotation. The scree tests for the two subsamples (Figures 1 and 2) display a notably similar pattern. However, they fail to indicate an obvious point of discontinuity to use as a cutoff for the number of factors to retain for rotation. 63 SCREE TEST Frankfort of Side "A" 7.0 5.0 '1 4.0-1 3.0 - Eigenvalues 2.0 --1 1.04 0.0 1 I r T 1 .— q Cd —1 Factor # Figure 1 -- Scree Test for Frankfort of Side "A" SCREE TEST Tawas of Side "A" 7.0 Eigenvalues 1 1 Factor # Figure 2 -- Scree Test for Tawas of Side "A" 64 Interpretability. The last method used to resolve the number of factors decision was interpretability. Runs were made specifying eight and nine factors for each sample. The results for the nine factor solution did not improve the interpretability of the rotated matrix; it only added a specific factor -- one on which only one variable loaded. In short, the results at eight factors were better than at nine. The four methods used to determine the number of factors to retain for rotation indicated.21 range between four and nine factors. At four factors only 42.92 and 42.62 of the total variation in the data would be explained for the Tawas and Frankfort samples respectively -- which is not very high. At nine factors 64.52 and 63.02 of the total variance would be explained. But, at nine factors, the eigenvalue for the Frankfort sample is .924, which is below the cutoff (H? 1.0. Further, at nine factors there is no substantial improvement from the perspective of interpreta- bility. These results led to the decision to retain eight factors from each subsample for rotation and subsequent comparison. Visual Comparisons The visual comparisons made across samples in this study are: (1) the number of factors with eigenvalues greater than one, (2) the percent of total variance explain- ed, (3) the configuration of factor loadings, (4) the 65 complexity of the factor structure, and (5) the communali— ties of the variables. The Number of Factors with Eigenvalues Greater Than 235. The previous section on the number of factors to extract focused on determining the number of factors to retain for rotation. This section explicitly compares the number of factors with eigenvalues greater than one (in the initial factor matrices) as an indicator of the similarity of factor structure across samples. As discussed in the previous section one of the "rules of thumb" in factor analysis is to extract and rotate only those factors with eigenvalues greater than one. This criterion for factor rotation is popular for its simplicity and relatively accurate performance. For the present study, the criterion for this indicator of factor structure similarity specifies that the number of factors with eigenvalues greater than one for the two matrices be within one. As displayed in Table 5, eight factors for the Frankfort sample and nine factors for the Tawas sample had eigenvalues greater than one. A difference which may be related to differences in respondents' familiarity with the target region. Obviously, the Tawas sample is not likely to be as familiar as the Frankfort sample with the northwestern coast of Lake Michigan -- Side "A". Thus, the Frankfort sample's fewer factors may indicate greater cohesion in the factor structure for that group in comparison with the Tawas 66 sample. The significance and magnitude of this difference, however, is difficult to assess. Is it an actual difference or just a random fldctuation in the data? There are no established methods to determine which is the case. Nevertheless, the number of factors with eigenvalues greater than one is within one for the two matrices. According to the established criterion, this finding provides an indica- tor of similarity for the factor structures for the two samples. Percent of Variance Explained. The percent of total variance explained was also considered 1J1 determining the number of factors to extract for rotation. The percent of total variance explained measures the relative importance of the factors in accounting for the relationships in the data. Here, the percent of variance explained by the two factor matrices prior to rotation serves as another indica- tor of the similarity of factor structure across samples. The criterion for this measure was established for both factor to factor comparisons and matrix to matrix comparisons. For individual factor to factor comparisons, the benchmarkis a +/- two percent difference between the two subsamples on.£1 given factor. Regarding the matrix to matrix comparison, the factor structure of tuna factor matrices is similar when (at the cutoff for factors to retain for rotation) the difference in the percent 01' total variance explained is +/- five percent between the two . 67 Comparing factor to factor, the percent of total variance explained by the two factor matrices is very similar. Moving from Factor 1 to Factor 27 one can see a notable pattern in both factor matrices (Table 5). The maximum difference between the percent of total variance explained for the two subsamples on a given pair of factors is .6 percent. This is well within the criterion range of +/- 2 percent. For the matrix to matrix comparison, at the cutoff point of eight factors (the number of factors retained for rotation),, the percent of total variance explained by the Frankfort sample is 59.6 percent (Table 5). For the Tawas sample it is 60.7 percent. The difference between the two, then, is 1.1 percent which is well within the standard for this measure of factor structure similarity. These findings increases our confidence 1J1 the stability of the factor structure of destination images across samples. It suggests that the underlying factor structure in the data is not specific to a particular subsample, but rather is inherent in the domain (of destina- tion images) under investigation. The following section continues this analysis by examining the rotated factor matrices: it begins by comparing the configuration of factor loadings across samples. Configuration of the Loading_. The varimax rotated factor matrices for the two samples on the twenty-seven regional image variables are presented separately in Tables 68 6 and 7; and together in Table 8. Factors for the Frankfort sample are labeled F1 to F8; for the Tawas sample they are labeled T1 to T8. To aid in assessing factor stability, the criterion pertinent to the configuration of factor loadings specifies that.£i similar factor across two matrices is one in which (at least) two variables load highly on that factor across matrices. A high loading is operationally defined as a variable loading greater than .38 on a factor. Summaries of the variables loading greater than .38 on the factors for the Frankfort sample are displayed from two perspectives in Tables 9 and 10. Table 9 presents the configuration of the high loading variables in tflue order displayed on the original factor matrix. Table 10 presents the same information, but ranks the loadings from highest to lowest for each factor. Factor F1 had high loadings on seven variables including: sandy, fun, enjoyable, colorful, appealing, outdoor-oriented, and interesting. The second factor (F2) loaded also loaded seven variables, which were: secluded, peaceful, unspoiled, delightful, quiet, natural, restful. Factor F3 had high loadings on three variables: friendly, family-oriented, and courteous. Factor F4 loaded with clean, pleasant, and scenic. The fifth factor for this sample (F5) contained high loadings on accessible and middle-class-oriented. The sixth factor (F6) loaded with forested and restful (note: forested had a positive loading, 69 mmmwo.31 mmmmo.o smm®3.¢l ooom:.o owmoo.on mcqmm.a m-oc.on o~a0¢.o oodeo.o mmoo~.o ~m~oe usuumou Hummus: mouse . concouuolumdusoa weuumououcu umusomuonuoouuso Masseussue «caucuses asuuomou vagueness . oncoom muamhonco msoouusou eoucoiuolamusmu csu xenon vou=ouuoummonutogeeaa acmmmoga ”amount; .nnvmonmu cosmouou . coao s>um=oaxo sovsmuon cacao ouauamouum u4= fiend I zv :<: seam uo magnum uuouxcoum not amuse: souomh cognac: wolquo> o munch 70 nsfioo.OI nmo~o.cl Omuno.o m~ao~.o ~mm~3.©l omna~.O ocmoo.au m~nm0.0 swam—.c mmao~.c ohmao.c ¢o¢~0.31 econ—.9 sqmoo.o nooem.o men—0.3 0c_n~.ol Name—.6 «dun—.OI menoo.: sonn~.o mafia—.o Onaso.o| m-mo.o| oo~m_.o numm~.o cacno.o aa~_o.o mqmmo.on omaoo.o mm—o~.o ~0qo_.o ammoc.a mm~_~.o oomo~.o mmomo.o: mcmm~.o Namoo.01 o~mc0.0n ammo—.0 Macao.o cecmo.c ~oo~o.on nooos.31 H~A anma I 2v :<: ovum no o—aemm smash mom uuuaoz souumm vououou umeuum> b manna 71. .uouumu a co mm.01\+ coca monsoon mauamou momamnuo> ououuuca monogamoumm ”ouoz e~.o mo.os o~.o Ann.oo oi.o oo.o A¢<.oo No.8 oo.o- o~.o Ho.ou no.o no.o -.ou eo.o. «s.o euaosusuu mo.ou mo.o o~.o -.o- ss.o Amn.o-o no. ~_.o Ann.oo a..o an.o o_.o o_.o Ams.oo .n.o e~.o «aunaot o~.o oo.os so.o -.o. no.o no.0 ha.ou mo.o Aoq.oo o~.o m~.o «n.o an.o Amq.oo mm.o -.o Hanson: oi.o- o~.o 5N.o mo.o- mo.o -.o. n~.o ~o.o Ano.co as.o o~.o o~.o no.o Amo.oo .o.o1 no.o sauna Aoo.oo oo.o- ii.o m~.o No.o mo.o c_.o so.o ~3.on ~_.o No.0 an.o mo.o so.o. ~o.o mo.o euuaustouuusuaon s~.o mm.o mo.o- 5N.o. so.o- ao.o so.o oo.on oo.o -.o so.o o_.o m~.o _~.c Aun.oo Ame.oo masnautuncs es.o to.o- s~.o no.o o~.o ~o.o mo.o -.o n~.o so.ou «2.0 am.o -.o a~.o so.o. Ao¢.oo euuaostoutooeuao ao.o Asq.ov Ho.o- s..o m~.o o~.o- so.o mo.o e~.o o~.o -.o -.o Aom.oo Ams.oa n~.c -.o saaugmssoe No.o- so.o ~o.o- n~.o Ame.oo no.on m~.o o~.o o~.o on.o em.o n~.o ao.o -.o «m.o Ao«.oo messaoaaa e~.o _~.o o~.o o_.o o~.o -.o so.oa so.oa n~.o n~.o eo.o n~.o oo.on e_.o Ano.oo Aus.oo flattened o_.o so.o so.o oo.on si.o n~.o so.ou q~.o «n.o oo.o- oo.o H~.o an.o Ann.oo k~.o m_.o eassoauaa mm.o q..on ~o.o oo.o- ca.o oo.o as.o n~.o a~.o Ans.oo so.o “N.o ao.o o~.o e~.o nn.¢ unseen No.o oo.on No.0. io.o «a.o s~.on Ak¢.oo e~.o oo.o ha.o -.c m~.o ~_.o. no.o Aen.oo Aeh.oo usaosofiao mH.o- oo.o so.o- ~o.o. ~o.o oo.o o~.o oo.o «a.o ¢~.o sm.o Ako.ov Anq.oo -.o o~.o -.c aaoantaou k~.o oo.o No.o- No.o. mo.o s~.o. o~.o N~.o -.o no.o Amq.oo Ahq.ov mm.o Ha.o m~.o «a.o eouaosoousss-au s_.o- mo.o ma.o no.o om.o mo.o so.o so.o so.o. so.o ks.o so.o o~.o mH.o Amn.oo “Ne.oo can oi.o n~.o o_.o oo.o: Anm.oo eo.o oo.o o~.o oo.o m~.o mo.o mo.o ~o.o ~_.o o~.o Ane.ov henna n~.o No.o es.o No.o ”H.o- -.o _~.o Aho.oo ¢~.o oo.o. n_.o n_.o «m.o n5.o N~.o n~.o concentouuaoau-oseusa «2.0 m~.c ~_.o ~_.o -.o No.ou no.ou no.on -.o Aon.ov As~.ov mo.o so.on e~.o e_.c ks.c uaaaoosa mo.o. oo.o. mm.o mo.o- «2.0 No.o- no.o n~.o «n.o sn.o Aon.ov ~o.o a~.o Ass.oo mo.c mo.o Macaques ao.ou k~.o n~.o _o.o so.o no.o no.o o~.o Ho.o «o.o Awm.ov Ao<.oo Aon.ov mo.o -.o o~.o , saunastu n~.o oo.o. oo.o so.o- so.o Ahe.ov no.o n~.o mo.o -.o mo.ou h~.o Ano.ov ao.o no.ou o~.o announce e~.o so.o Ao~.oo o_.o no.o so.o: so.ou n~.o mo.o ni.o «s.o N~.o o~.o. o~.c ~n.o oo.o some mo.o s~.o mo.o- Aon.oo no.o Ho.o o~.o Aie.oo- mo.ou k~.ou wo.ou oo.on c~.o. oo.on -.ou os.ou o>saaoaxo so.o mo.o Anq.oo Ho.o. so.o oo.o- so.o so.oa s~.o ~o.o no.o. _~.ou -.o Am<.oo oo.ou n~.o eoeasuoa mo.o- _a.o m~.o N~.o mo.o H_.o mm.o oo.c m~.o ion.oo so.o m~.o ~_.c is.o A_e.co k~.o cause o~.o «n.o -.o. eo.o. so.o e_.ou Ano.oo ak<.oo oo.o -.o ~o.o mu.o ao.o «~.o. ao.c oo.o eonsaaouuu we we as he as on we no. as «a me am as «a "a as .:.. u... .uuu..|.a: u5m m moso ~~---a vbile 38Ven1 and de ble: 72 Table 9 Configuration of the Highest Loadings Variables for Frankfort Sample of Side "A" (N = 134) VARIABLE F1 F2 F3 F4 F5 F6 F7 F8 accessible 0.47 clean 0.39 secluded 0.48 expensive 0.39 open forested 0.47 friendly 0.40 peaceful 0.44 pleasant 0.59 middle-class- oriented 0.67 sandy 0.43 fun 0.62 family-oriented 0.47 courteous 0.67 enjoyable 0.74 scenic 0.45 unspoiled 0.55 colorful .42 appealing .46 delightful outdoor-oriented interesting tourist-oriented quiet 0.63 natural 0.45 restful 0.42 -0.54 developed 0.51 .46 .43 CO 00 while restful had a negative loading on this factor). The seventh factor, Factor F7 loaded two variables: expensive and developed. Finally, Factor F8 loaded a single varia- ble: delightful. 73 Table 10 Highest Loading Variables Ranked by Factor for the Frankfort Sample of Side "A" (N = 134) VARIABLE F1 F2 F3 F4 F5 F6 F7 F8 enjoyable fun appealing outdoor-oriented interesting sandy colorful quiet unspoiled delightful secluded natural peaceful , restful .42 courteous 0.67 family-oriented 0.47 friendly 0.40 pleasant ~0.59 scenic 0.45 clean 0.39 middle-class- oriented 0.67 accessible 0.47 forested 0.47 restful -0.55 developed 0.51 ex ensive 0.39 de ightful 0.47 0000000 b 0‘ .63 .55 .48 .48 .45 0000000 The high loading variables for the Tawas sample's factor matrix are similarly summarized in Tables 11 and 12. The first factor in that matrix, Factor T1, loaded the following variables: clean, fun, enjoyable, colorful, and interesting. Factor T2 loaded with the following: for— ested, friendly, courteous, and delightful. The third factor (T3) contained high loadings on: peaceful, pleasant, 74 Table 11 Configuration of the Highest Loadings Variables for Tawas Sample of Side "A" (N = 125) VARIABLE T1 T2 T3 T4 T5 T6 T7 T8 accessible 0.65 clean 0.41 secluded 0.45 expensive open 0.70 forested 0.65 friendly 0.50 peaceful 0.59 pleasant 0.74 middle-class- oriented sandy 0.85 fun 0.52 family-oriented 0.43 courteous 0.45 enjoyable 0.56 0.47 scenic unspoiled colorful 0.63 appealing 0.42 delightful' 0.39 outdoor-oriented interesting 0.52 tourist-oriented 0.69 quiet 0.65 natural 0.49 restful 0.57 . developed 0.49 and family-oriented. Factor four, or T4, loaded highly with the following: quiet, natural, and restful. Three variables loaded on the fifth factor for this sample (T5): accessible, enjoyable, and developed. Factor'fr6 had high loadings on: sandy and appealing; and the seventh factor T7 had high loadings on: secluded and open. Finally, the last factor for this sample (T8) loaded only a single variable: tourist-oriented. 6: t1: 75 Table 12 Highest Loading Variables Ranked by Factor for the Tawas Sample of Side "A" (N a 125) .‘-...=....=‘==-8.===.=$B==8:==============B====338828===== VARIABLE T1 T2 T3 T4 T5 T6 T7 T8 colorful 0 enjoyable 0 fun 0.52 interesting 0 clean 0 forested 0.65 friendly 0.50 courteous 0.45 delightful 0.39 pleasant 0.74 peaceful 0.59 family-oriented 0.43 quiet 0.65 restful 0.57 natural 0.49 accessible 0.65 developed 0.49 enjoyable 0.47 sandy 0.85 appealing 0.42 open 0.70 secluded 0.45 tourist-oriented . 0.69 A comparison of the configuration of the high loading variables on the factors indicates both similarities and differences. The first factor for the Frankfort sample (F1) includes high loadings on the following variables: sandy, fun, enjoyable, colorful, appealing, outdoor- oriented, and interesting. Four of these variables are also found in the first factor for the Tawas sample (T1): fun, enjoyable, colorful, and interesting. In addition, two of the variables loading highly on Factor F1 not found on 76 Factor T1 are found together on Factor T6: sandy and appealing. The second factor for the Frankfort sample (F2) includes seven high loading variables: secluded, peaceful, unspoiled, delightful, quiet, natural, and restful. Three of these variables load together on Factor lflizfor the Tawas sample: quiet, natural, and restful. The remaining variables from Factor F2, however, do not load together on a factor for the Tawas sample; they are found on factors mixed in with other variables -- secluded on T7, peaceful on T3, and delightful on T2 (the remaining variable, unspoiled, did not load highly on any of the factors for the Tawas sample). The third factor from the Frankfort sample (F3) was the only other factor to show similarity with the Tawas sample. Factor F3 included high loadings on the following variables: friendly, family-oriented, and courteous. Two of those variables are found loading highly together on Factor T2: friendly and courteous. However, two other varia— bles: forested and delightful also loaded on Factor T2 -- adding some confusion to this dimension for this sample. The third variable from Factor F3, family-oriented, loaded on Factor T3 along with the variables pleasant and peace- ful. These three variables are somewhat intuitively related to Factor F3, but are not supported by a comparison of the configuration of high loading variables. 77 No other factors between the two matrices included variables which loaded together consistently. Either variables which loaded together on a factor for one sample -- such as Factor F4 -- were split between factors for the other sample, or they appeared together on a factor for one sample and not at all for the other sample -- such as F7. This comparison indicates that three of the eight dimensions show some degree of visual similarity across samples -- Factor F1 with Factors T1 and T6, Factor F2 with T4, and Factor F3 with T2. The next visual comparative aspect to be discussed is factor complexity. Complexity. As discussed in the literature review, although implicitly involved in the configuration of the variables, the complexity of the variables which load on a set of factors or the complexity of the factors themselves, may be compared explicitly. This section examines both aspects of complexity. The first is the complexity of the variables -- on how many factors does a variable load highly. The second aspect considered in this section is the complexity of the factors. This measure of comparability looks to see what happens to a factor from one sample to another. Specifically, does a factor in one matrix split into two or more factors in another. The criterion for comparing the complexity of the variables between matrices specifies that a variable with the same complexity on two factor matrices indicates stabi- fa We th Ta‘ Tax 78 lity for the factors involved with those variables. An operational indicator of a: stable dimension between matri— ces, the other measure of complexity, is when a factor found in one matrix can be identified in one or more of the factors in the other matrix. Regarding the complexity of the variables, for the Frankfort sample two of the twenty seven variables loaded highly (n1 more than one fattor. The variable delightful loaded on Factors F2 and F8; and variable restful loaded on Factors F2 and F6. For the Tawas sample only one variable loaded on more than one factor -- enjoyable; which loaded on Factors T1 and T5. All other variables loaded, if at all, on only one factor. Thus, with only a few excep— tions the majority of the variables loaded on a single factor, and those which did load on more than one factor were not the same variables for the two samples. For the complexity of the factors, one factor from the Frankfort sample which split between factors in the Tawas sample is Factor F1. This factor can be found in the Tawas matrix in Factors T1 and T6. Another factor from the Frankfort sample, F2, can be found, for the most part, in Factor T4, however, other variables in F2 are found in Factors T2, T3, and T7. In addition, Factor F3 from the Frankfort sample split between Factors T2 and T3 in Tawas sample. Both Factors T2 and T3, however, also have other variables which did not load highly in Factor F3 (Factor T3 has only one variable in common -- family—oriented -- with 79 Factor F3). This indicates that some "noise" itsxnixed in with this dimension for the Tawas sample. Each sample had a specific factor -— one (”I which only a single variable loaded. However, the specific factor did not involve the same variables for time two matrices. The two factors were: for the Frankfort sample, F8 -— with the variable delightful —- and for the Tawas sample, T8 -— with the variable tourist-oriented. In summary, an examination of the complexity of the variables involved with each factor matrix and the complex— ity of the factors between matrices provided only a slight increase iJl our understanding of the factor structure for the two samples. The analysis centering on the complexity of the factors lent more insight into the stability of factors, than did the analysis of the complexity of the individual variables. For the most part, the more positive results from this comparison affirmed the findings from the previous comparison involving the configuration of factor loadings. Communality of the Variables. The final visual comparison conducted in the comparison across samples entails an examination of the communalities of the variables following factor rotation. The communality of a: variable is a measure of the portion of a variable's variance which is accounted for by the factors extracted. Communalities are calculated by summing the squares of the factor loadings for a row (that is, for a given variable) in the factor 80 loadings matrix. A comparison of a variable's communality between studies -- in this case, between samples —— helps distinguish between those variables that are consistently interrelated and those that are consistently unique. Communality comparisons are not frequently employed in factor comparison studies. The literature reviewed regarding this area of factor comparison only suggested that a variable's communality across two studies could be compared. No literature was found which provided any specific guidance regarding the criteria used to assess the stability of a variable's communality across studies. To establish criteria for this section, each variable for the two samples was assigned to one of three categories depending on the ranking of the variable's communality relative to the other variables. The three categories were (1) the upper third (which contains the variables with the top 9 communalities), (2) the middle third (with the variables having the next 9 highest communa- lities), and (3) the lower third (the variables with the 9 lowest communalities). Stability is indicated when a varia- ble remains in the same communality category (or third) for both samples. Those variables which remain in the top third between samples provide an indication of variables which are consistently intercorrelated with (and important to) the factor structure underlying the data. Table 13 presents the communalities for each variable by subsample, and Table 14 shows the communality' I‘.Jlllc-nrd§* 81 category that each variable was classified into by sub- sample. Only four variables remained in the first category (the top third) between samples. The four variables are: enjoyable, pleasant, quiet, and restful. The variables consistently found in the second category (the middle third) between the Table 13 Communalities By Subsamp1e* COMMUNALITY ABSOLUTE ----------------- DIFFERENCE VARIABLE FRANKFORT TAWAS IN PERCENT accessible 55.92 47.42 8.52 clean 33.82 42.92 9.12 secluded 26.72 28.02 1.22 expensive 40.12 16.32 23.82 open 17.52 64.82 47.22 forested 42.62 46.62 4.02 friendly 29.82 43.72 14.02 peaceful 33.02 68.32 35.42 pleasant 49.12 66.52 17.42 middle-class-oriented 52.32 28.92 23.42 sandy 30.92 78.72 47.82 fun 42.22 47.92 5.72 family-oriented 32.12 44.02 11.92 courteous 54.12 46.52 7.62 enjoyable 69.62 61.72 7.92 scenic 47.72 26.92 20.82 unspoiled 45.82 29.32 16.52 colorful 42.42 47.82 5.42 appealing 48.72 51.82 3.12 delightful 67.32 38.02 29.22 outdoor-oriented 40.02 19.72 20.32 interesting 52.42 38.12 14.32 tourist-oriented 21.82 53.72 31.92 quiet 51.62 56.42 4.82 natural 46.02 53.92 7.92 restful 64.22 56.32 7.82 developed 30.02 31.92 1.92 * Communalities reported are for the rotated. matrices. 82 Table 14 Variables Classified into Communality Thirds (Upper, Middle, and Lower) by Subsample FRANKFORT TAWAS * enjoyable sandy delightful peaceful * restful pleasant accessible open Ema: courteous enjoyable a; interesting quiet ::E~ middle-class-oriented restful * quiet natural * pleasant- tourist—oriented * appealing appealing scenic fun natural colorful Egg unspoiled accessible gfi * forested forested H: * colorful courteous 2'“ * fun family-oriented expensive friendly outdoor-oriented clean clean interesting peaceful delightful family—oriented developed a: a: sandy unspoiled mo: * developed middle-class-oriented '3'; friendly secluded .qe« * secluded scenic tourist—oriented outdoor-oriented open expensive Indicates variables remaining in the same third across samples. two samples are: appealing, colorful, forested, and fun. And finally, the two variables which remained in the bottom or lowest third category in both samples are: developed and secluded. All the variables which remained stable in the first two categories in this comparison were also among those 83 found to indicate stability in the previous comparison involving the configuration of factor loadings. Although this is not surprising (since the communalities are computed from tflua same factor loadings), this finding provides additional information regarding the stability of the factor structure underlying destination images. The following discussion centers on the mathematical approaches used to assess factor similarity, vector comparisons. Vector Comparisons Generally, vector comparisons provide more objective measures of factor similarity than do the visual compari— sons. These approaches compare the factors from one matrix with the factors from a second. The vector comparisons applied in this phase are: (1) Pearson's correlation coeffi- cient (r), (2) root mean square (RMS), (3) coefficient of congruence (CC), and, (4), the salient variable similarity index (S-Index). In all cases the object of comparison used in calculating these measures was the factor loadings. Pearson's Correlation Coefficient (r). Pearson's product moment correlation coefficient is a measure of pattern similarity between two vectors. The criterion pertinent to this measure of factor similarity, considers two factors from separate matrices similar (that is, picking up the same underlying dimension) if: (1) the two factors correlate above .40, and (2) that relationship is statisti- cally significant (at the Prob < .05 level). 84 Each of the eight factors from the Frankfort sample was paired with each of the eight factors from tflua Tawas sample. Pearson's correlation coefficient was then calcu- lated for each factor pair. As seen in Table 15, several factors displayed significant relationships across the two samples. In particular, Factors F1 auui'Tl, Factors F1 and T6, and Factors F2 and T4 all correlated significantly above the criterion of .40. Factors F3 and T2 also displayed a Table 15 Factor Comparison Across Samples Using the Correlation (Pearson's r) of Factor Loadings F1 F2 F3 F4 F5 F6 F7 F8 T1 0.68 T2 0.01 T3 0.01 (.001) (.977) (.957) 0.09 (.665) 0.28 (.157) -o.1o (.158) 0.11 (.569) -0.02 (.910) -0.33 (0096) 0.20 (.308) 0.22 (.274) 0.39 (.044) -0.12 (.566) 0.33 (.090) 0.31 (.110) -0.50 (.008) 0.06 0.28 (.159) 0.19 (.337) 0.36 (.064) 0.09 (.657) -0.38 (.052) -0.09 (.662) —0.03 (.877) T4 -0.10 (.622) 0.75 (.001) 0.08 (.706) 0.23 (.245) 0.15 (.457) —O.34 (.079) -0.51 (.006) -0.16 (.417) T5 0.09 (.649) -0.60 (.001) 0.12 (.540) 0.02 (.902) 0.32 (.105) -0.24 (.236) 0.29 (.145) -0.09 (.662) T6 0.47 (.014) 0.02 (.924) -0.18 (.357) 0.20 (.326) -0.10 (.631) -0.03 (.879) 0.12 (.568) 0.07 (.728) T7 -0.22 (.260) 0.29 (.149) -0.35 (.074) -0.02 (.941) 0.08 (.680) 0.00 (.984) 0.14 (.496) -0.31 (.121) T8 -0.15 (.462) -0.24 (.228) 0.09 (.644) 0.03 (.885) -0.02 (.934) 0.21 (.286) 0.14 (.474) -0.20 (.325) Note: Factors F1 to F8 are from the Frankfort sample. Factors T1 to T8 are from the Tawas sample. Parentheses indicate significance level. '85 significant relationship, but correlated at: .39, just slightly below the criterion of .40. Other notable correlations were obtained, but they were not at the .05 level of significance and several were inverse relation- ships. This analysis indicated that the pattern of factor loadings for several of the factors was very similar. In fact, the factors found similar in this comparison were the same pairings identified in time comparison involving the configuration of factor loadings. Since this particular analysis is only sensitive to pattern similarities, other factor comparison nmasures were calculated to assess both pattern and magnitude similarities of the factors across the two samples. Root Mean Square (RMS). The Root Mean Square or RMS provides the strictest measure of factor similarity since variations in both pattern and magnitude are detected. RMS ranges from zero (for a perfect pattern-magnitude match) to two (for a perfect inverse match). Intermediate values of RMS, however, are not readily interpretable. For the present study, RMS coefficients less than or equal to .10 provide a stringent measure of the match between a pair of factors. RMS was calculated for all combinations of factor pairs across the two samples (Table 16). The coefficients were, in general, quite low -- ranging in value from a low of .14 to a high of .35. None, however, were low enough to 86 Table 16 Factor Comparison Across Samples Using the Root Mean Square (RMS) of Factor Loadings T1 T2 T3 T4 T3 T6 T7 T8 F1 0.17 0.28 0.28 0.29 0.27 0.22 0.31 0.31 F2 0.28 0.25 0.24 0.14 0.35 0.28 0.24 0.31 F3 0.23 .20 0.23 0.24 0.24 0.27 0.29 0.25 F4 0.23 .27 0.20 .21 0.24 .22 .25 .25 F5 0.28 .23 0.26 .25 0.21 .27 .24 .24 .34 0.30 .33 0.22 .24 .24 .22 0 0 0 0 0 0 0 0 0 0 0 F6 0.34 0.28 0.35 0.33 0.30 0.28 0.27 0.23 F7 0.35 0 0 0 0 0 0 0 0 0 0 F8 0.26 .25 0.26 .26 0.22 .22 .26 .24 Note: Factors F1 to F8 are from the Frankfort sample. Factors T1 to T8 are from the Tawas sample. meet the established criterion of .10, although two factor- pairs did have coefficients under .20: F1 with T1 (.17) and F2 with T4 (.14). These two pairs of factors were among those found to be similar in the previous comparisons. The RMS coefficients for other notable factor pairs were:.22 for F1 and T6; and .20 for F3 and T2. Both of these factor pairs were also found to display a degree of similar— ity between samples in other comparisons, but did not meet the standard for this, more rigid, comparative measure. Although the RMS coefficients calculated for the factor pairs found similar in the previous comparisons did not meet the established cutoff level, the coefficients for the factors involved in those pairs displayed the lowest 87 (relative) coefficients of the factors involved in those pairings. For example, for Factor F2 the lowest RMS for the Tawas factors calculated with Factor F2, was for Factor T4. The same was true for Factor T4 -- the lowest of the eight coefficients calculated using T4 was for Factor F2. A number'9H4 Amm~ I 2V :<: ovum no museum vocuanoo mom sauna: wouuom soumuom umauum> «a ounce 1(14 Ocomm.o mmNao.oi eoNNN.ol oNNmo.0i o—om~.o Ncnwo.01 quon.OI Gasma.c o—mm~.o ooNoo.OI mohao.:l ecoNo.OI n~qoo.oi mammo.o nNQNo.o oqoqo.oi mmq—~.o QNNa~.o nhooc.01 Nonoo.:i cmnNN.OI cch—.o Nc-~.o anooo.o nmmo~.o o~n00.01 nose—.3 thqov gsuumou Housumc seas: coucowuoium«s=ou mcwumououcn concouuoiuooeuso "sangussoe newnmoaam Hauuodou vomaoamcs oacoom ounmaofico usomuusou euucmuuoiaauaou csu mecca coucoquouoomuuioaceua acommoma asuoumon issuance noumouou coco usumcoaao cuesHuou manna nuanmmouom uan Awm_ I 2v :n: ovum no museum cocuasoo com swans: nouoam mousse: usnauc> MN munch 1()S .uouumu o co mm.o:\+ coca uuumoum mcqomoa moaamwum> oumuuocw monogamoumm .uouoz 500.00 o~.0: 00.0: 0~.0 5N.0: ~N.0 00.0: 00.0: 0N.0 00.0: 00.0: n~.0 o~.ol 0~.0 no.0 MN.0: 0~.0 50.0 0~.0: 0~.o ~0.0: o~.o no.0: n~.0 00.0: 50.0 no.0 00.0 50.0 00.0 00.0: H0.0 -.0 “0.0: o~.0 ~—.0 00.0:. "0.0: 00.0: 00.0 MN.0I 00.0 N~.0 0N.0 -.0 50.0 0~.0 550.0:0 5_.0 00.0 50.0: -.0 0~.0 00.0 50 5< 5 m050ov usuuoou amususc uoqso voucofiuoiumwuoou ocuuoououcn voucoquoluooouso Hoongosooo ucfifloooos usuuoaou voaaooncs uwcouo ounoaonao ssoouusoo nonconoolaawlmu mam 50sec usuaowu0:mmm~u:mu00«a ucommono asuoumoo haccuuuw wouoouou cooo o>wmcaoxo venomous condo uncommouuo mam<~u<> :m: a :<: mouum soon we odonom vocdoaou com sauna: uouuam unusuom xmswus> 0N munch 106 Side "B". The two matrices are also presented together, factor by factor, in Table 24. In that table, factors for the adjectives describing Side "A" are labeled A1 to A7, and the factors for the adjectives describing Side "B" are labeled B1 to B7. A5 in the previous comparison, a stable factor is one on which (at least) two variables load highly (above .38) together on that factor across matrices. The configu- ration of each factor matrix will be presented first, followed by a discussion of the similarities and differences between the two. As seen in Tables 25 and 26, the first factor for the factor matrix summarizing respondents' images of Side "A" (Factor A1) contained the following high loading varia- bles: forested, friendly, peaceful, middle-class-oriented, family-oriented, courteous, and delightful. Factor 2 for the Side "A" matrix (A2) included six high loading varia- bles: secluded, peaceful, unspoiled, quiet, natural, and restful. Factor A3 had high loadings on these adjectives: fun, enjoyable, colorful, delightful, and interesting. The fourth factor for this matrix (Factor A4) loaded with the following variables: peaceful, pleasant, sandy, enjoyable, and appealing. Factor A5 loaded two variables: accessible and enjoyable. Finally, the last two factors for this matrix each loaded a single variable. The variable tourist- oriented loaded on Factor A6; and the variable expensive loaded on Factor A7 (with a negative loading). for Combined Sample of Side "A" (N VARIABLE accessible clean secluded expensive open forested friendly peaceful pleasant middle-class- oriented sandy fun family-oriented courteous enjoyable scenic unspoiled colorful appealing delightful outdoor-oriented interesting tourist-oriented quiet natural restful developed A1 0.41 0.58 0.38 107 Table 25 Configuration of the Highest Loading Variables A2 0.40 0.41 0.72 0.50 0.56 A3 0.61 0.48 0.58 A4 = 188) A5 A6 A7 0.62 -0.67 0.54 ' 108 Table 26 Highest Loading Variables Ranked by Factor for the Combined Sample of Side "A" (N a 188) VARIABLE family-oriented friendly courteous middle-class- oriented delightful forested peaceful quiet restful natural peaceful unspoiled secluded colorful interesting delightful fun enjoyable pleasant sandy appealing peaceful enjoyable accessible enjoyable tourist-oriented expensive A1 A2 0.72 0.56 0.50 0.46 0.41 0.40 A3 0.61 0.58 0.48 0.40 0.38 A4 0.51 0.48 0.48 0.44 0.40 A5 A6 A7 The variables loading on the first factor for Side "B" (B1) (see Tables 27 and 28) were: secluded, open, peace- ful, Factor B2 pleasant, included high junspoiled, family-oriented, enjoyable, ing. bles: sandy, fun, colorful, and delightful. The third factor (B3) quiet, loadings natural, and 011 loaded highly on restful. the following: fun, outdoor-oriented, and interest- four varia- Factor B4 109 Table 27 Configuration of the Highest Loading Variables for Combined Sample of Side "B" (N a 188) VARIABLE Bl B2 B3 B4 B5 B6 B7 accessible 0.43 clean secluded 0.43 expensive 0.90 open 0.46 forested 0.70 friendly 0.53 peaceful 0.48 pleasant 0.42 middle-class- oriented 0.52 sandy 0.48 fun 0.44 0.44 family—oriented 0.47 courteous 0.54 enjoyable 0.50 scenic unspoiled 0.48 colorful 0.60 appealing delightful 0.52 outdoor-oriented 0.52 interesting 0.40 tourist-oriented quiet 0.77 natural 0.53 O restful .51 developed 0.60 loaded the following: accessible, friendly, middle-class- oriented, and courteous. The remaining three factors for this matrix each contained only one high loading variable. The variable expensive loaded on Factor BS; forested loaded on B6; anddeveloped loaded on Factor B7. A comparison of the configuration of the high loading variables on the factors shows some similarity between the two matrices. The first factor for target 110 Table 28 Highest Loading Variables Ranked by Factor for the Combined Sample of Side "B" (N = 188) VARIABLE B1 B2 B3 B4 B5 B6 B7 quiet natural restful unspoiled peaceful open secluded pleasant outdoor-oriented enjoyable family—oriented fun .44 interesting .40 colorful 0.60 delightful 0.52 sandy 0.48 fun 0.44 courteous 0.54 friendly 0.53 middle-class-oriented 0.52 accessible 0.43 expensive 0.90 forested 0.70 developed 0.60 00000000 b on .52 .50 .47 00000 region "A" (A1) contained high loadings on the following variables: forested, friendly, peaceful, middle-class- oriented, family-oriented, courteous, and delightful. Three of these variables are also found for Side "B" on Factor B4: friendly, middle-class-oriented, and courteous. None of the other four variables on Factor A1 are found together on a factor for target region "B". The second factor for Side "A" (A2) included six high loading variables: secluded, peaceful, unspoiled, quiet, natural, and restful. All six of these variables 111 were also found together on Factor (B1) indicating a high level of convergence for this dimension. Two variables ‘which also loaded on Factor B1 include: open, and pleasant. “The adjective open is not found on the factor loadings Inatrix for Side "A", however, the adjective pleasant is. IPleasant loads on Factor A4 along with four other varia- bles: sandy, appealing, peaceful (which was already found on IFactor A2), and enjoyable. Thus, of the variables found on 331, six are found together on A2 and two are found together (on A4 (though the two varibles on A4 also load with other 'variables not found on Factor B1). Factor 3 for Side "A" (A3) included the following adjectives: fun, enjoyable, colorful, delightful, and interesting. These same adjectives appear on the matrix for Side "B" in Factors B2 and B3. Of these six adjectives, three were found on Factor B2: fun, enjoyable, and interest— ing; and three were found on B3: fun, colorful, and delight- ful (note that the adjective fun loaded on two factors for Side "B" -- B2 and BB). The remaining factors did not involve results which were comparable between the matrices for the two target regions. To recap, the following factors for target region "A", were found to be similar with the following factors for target region "B": Factor A1 with B4; Factor A2 and B1; Factor A3 with Factors BZ and B3; and Factor A4 with Factor B1. The results indicate that there is some degree of similarity in the factor structure for the two target 112 regions. The next aspect of factor comparison considered is factor complexity. Complexity. As before, this analysis considers both aspects of complexity —- the complexity of the variables on the factors between studies and the complexity of the factors themselves between the two studies. The same criteria used in the comparison across samples are sought for this comparison: (1) for the complexity of the varia- bles, look for a variable loading highly on the same number of factors between studies, and, (2) for the complexity of the factors, look to see whether a factor found in one study can be clearly identified in one or more factors in another study. That is, did a dimension which was found in a single factor on one matrix remain as a single factor, or was it split between two or more factors, on the second matrix. Regarding the complexity of the variables, for the adjectives used to describing Side "A", three loaded on more than one factor in the Side "A" matrix (Table 25): peaceful, enjoyable, and delightful. The adjective peaceful loaded on three factors: A1, A2, and A4. The adjective enjoyable also loaded on three factors: A3, A4, and A5. Finally, the adjective delightful loaded on two factors: A1 and A3. The “adjectives involved on the factors for Side "B" did not display the same level of complexity as the Side "A" adjectives. For the Side "B" matrix, only one adjective loaded on more than one factor: fun -— which was found on Factors BZ and B3. In short, the adjectives involved in the 113 factor analysis for Side "B" were less complex than those involved in the factor analysis for Side "A". This may mean that respondents' had a more concise image —- and thus, a clearer simple structure —- of Side "B" than they did of Side "A". In the comparison of the complexity of the factors, of the factors found to be similar between the two matrices, only Factors A1 and B4 remained as one relatively identifi— able factor for both matrices. Otherwise, one factor from each matrix appeared as two factors on the other matrix. IFor Side "A", Factor A3 was found split between two factors (B2 and B3) in the Side "B" matrix. For the Side "B" factor Inatrix, Factor B1 was found in Factors A2 and A4 for the ESide "A" matrix. Thus, for this comparison, of the dimen- esions which appear stable, only one factor remained as as a asingle factor across matrices. The others appeared in more t:han one factor on the second matrix. In summary, the comparison of the complexity of the ‘fariables between the factor matrices for the two target Itegions did not provide any substantive support regarding tZhe stability of factor structure across regions. In fact, :1-t: demonstrated that the variables on the Side "B" matrix we re more often alligned with a single factor than were the ‘w’£3.riables for the Side "A" matrix. The comparison involving t he. complexity of the factors, on the other hand, did :IL'Jtl-dicate stability for several factors which retained lbi-eir identity in one or more factors across the two 114 matrices. The next portion of this analysis considers the communalities of the variables as the object of comparison between the two factor matrices. Communalipy of the Variables. This comparison, involving the communalities of the variables, looks to see whether a variable remains in the same communality category (or third) across studies. Once again, of particular interest are those variables which remain in the upper communality third in both factor matrices. The communalities for each region are presented in Table 29, while Table 30 shows the communality category that each variable was assigned to, for each region. Four variables were found in the upper communality third for both regions: expensive, colorful, quiet, and natural. For the middle category, two variables remained stable: unspoil- ed and interesting. Finally, four variables maintained their status as members of the lower communality third for both factor matrices: secluded, open, scenic, and tourist- oriented. Three of the variables found stable in the first two ctommunality categories (colorful, quiet, natural, unspoiled, aind interesting) were also among those determined to be £3table dimensions in the comparison of the configuration of factor loadings. One other variable, expensive, also JE‘Iemained in the upper third for both regions. However, it, ‘“-xalike the other variables found stable in this comparison, 115 Table 29 Communalities By Region* COMMUNALITY ABSOLUTE ----------------- DIFFERENCE VARIABLE SIDE "A" SIDE "B" IN PERCENT accessible 45.82 22.92 22.92 clean 35.32 35.52 , 0.22 secluded 17.62 28.02 10.42 expensive 54.32 81.82 27.62 open 14.42 29.42 15.02 forested » 35.82 56.32 20.62 friendly 38.92 47.12 8.22 peaceful 56.62 31.82 24.82 pleasant 42.72 31.22 11.52 middle-class-oriented 31.22 33.82 2.62 sandy 40.82 28.72 12.02 fun 36.82 44.72 7.82 family—oriented 47.52 32.52 15.02 courteous 47.42 41.92 5.62 enjoyable 57.62 40.92 16.72 scenic 34.22 20.42 13.82 unspoiled 36.82 34.82 2.02 colorful 50.42 44.02 6.42 -appealing 50.62 36.72 13.92 delightful 58.82 41.52 17.42 outdoor-oriented 25.22 43.92 18.72 interesting 44.62 32.52 12.02 tourist-oriented 32.22 26.92 5.42 quiet 61.22 64.12 2.92 natural 53.12 56.02 3.02 restful 60.32 41.32 18.92 developed 20.72 39.82 19.12 * Communalities reported are for the rotated matrices. 116 Table 30 Variables Classified into Communality Thirds (Upper, Middle, and Lower) by Region SIDE "A" SIDE "B" quiet expensive restful quiet delightful forested an enjoyable natural 3:155 peaceful friendly 9.3: expensive fun 3" natural colorful appealing outdoor-oriented colorful courteous family-oriented delightful courteous restful accessible enjoyable So interesting developed on: pleasant appealing ops H: sandy clean 2?” friendly unspoiled fun middle-class-oriented unspoiled interesting forested family-oriented clean peaceful scenic pleasant 6:0 tourist-oriented open "SE middle-class-oriented sandy SE outdoor-oriented secluded developed tourist-oriented secluded accessible open scenic * Indicates variables appearing in the same third across regions. appeared as a specific factor (one' on which only one Ivariable loaded highly) in both loadings matrices. of the factor structure of regional images remain stable across These findings target regions. The support the contention that elements following major heading is 117 concerned with the vector comparison approaches to assessing the stability of factors across regions. Vector Comparisons The vector comparison techniques applied to compare factors across regions generally parrallel those used in the comparison across samples, with one exception. 'Since this comparison involves the same subjects and different sets of independent variables, in addition to correlating the factor loadings, the respondents' factor scores were also correlat- Ed across the two studies. Correlating factor scores is recommended in this case because similarity of the factor loaidings does not guarantee similarity of factor scores in all cases. In summary, five vector comparison techniques will be used to compare factors across regions: (1) the Correlation of factor loadings, (2) the correlation of factor scores, (3) the root mean square (RMS), (4) the Coefficient of congruence (CC), and (5) the salient variable Similarity index (S-Index). For the root mean square, the Coefficient of congruence, and the S-Index the object of Comparison was the factor loadings matrices. 118 Pearson's Correlation Coefficient (r) of Factor Inmadings. Correlation coefficients were calcualted for each fax:tor for Side "A" (A1 to A7) paired with each factor fcxr Side "B" (B1 to B7) (Table 31). A number of factor [Hairs displayed correlation coefficients above .40: A1 aivd B4 (.63), A2 and B1 (.87), and A3 and B3 (.62). Each of thtase correlations were highly significant (at the Prob. < .05 level). Other significant factor pairs were: A3 anti B2 (.42) and A4 and Bl (.39). All of these factor pairs Table 31 Factor Comparison Across Regions Using the Correlation (Pearson's r) of Factor Loadings B1 B2 B3 B4 B5 B6 B7 A1 0.00 0.32 0.05 0.63 —0.39 0.20 —0.31 (.987) (.109) (.808) (.001) (.043) (.307) (.111) A2 0.87 0.14 -0.08 -0.31 -0.35 0.05 -0.40 (.001) (.109) (.675) (.121) (.075) (.790) (.041) A3 0.16 0.42 0.62 -0.14 -0.23 0.01 -0.45 (.421) (.028) (.001) (.479) (.243) (.949) (.019) A4 0.39 0.33 0.52 -0.12 -0.35 -0.49 -0.31 (.047) (.097) (.006) (.546) (.071) (.010) (.121) A5 -0.29 0.25 -0.17 0.39 -0.15 -0.28 0.20 (.137) (.215) (.388) (.046) (.466) (.157) (.326) A6 -0.55 -0.18 0.01 -O.13 0.03 0.40 0.45 (.003) (.381) (.961) (.534) (.893) (.037) (.019) A7 0.25 0.36 0.04 0.18 -0.72 0.37 -0.35 (.211) (.062) (.852) (.379) (.001) (.059) (.071) Note: Factors A1 to A8 are from Side "A". Factors B1 to B8 are from Side "B". Parentheses indicate the significance level. 119 were also determined to be similar in the section on the visual approaches used to evaluate factor structure across regions. Although there were other pairs of factors which displayed correlations that were significant at the .05 level, most were inverse relationships. The next section of this analysis considers the correlation of respondents' factor scores across regions. Pearson's Correlation Coefficient (r) of Factor Scores. As in the preceeding comparison, correlations were Computed for all the possible combinations of factors across the two regions. This time, however, factor scores for each respondent were used as the object of comparison. As seen in Table 32, the correlations of factor scores are generally lower than those based on the factor 10£idings; the highest coefficient calculated was .31 (for factor pair A2 and B1). Thus, none of the factor pairs met the cutoff level of .40 for this measure. Nevertheless, Several factor pairs did display relatively notable coeffi— cients for this comparison, they include: A1 with B4 (.23), 12 with Bl (.31), A2 with 132 (.22), A4 with 131 (.24), and A5 with B4 (.29). The majority of these pairs were 8“long those found to be similar in the previous compari- sons, The exceptions include: A2 with B2 and A5 with B4. An inspection of the loadings for these two factors shows that their similarity lies in the number of near-zero 120 Table 32 Factor Comparison Across Regions Using the Correlation (Pearson's r) of Factor Scores B1 B2 B3 B4 B5 B6 B7 A1 0.06 0.13 0.04 0.23 0.01 0.14 0.06 (.401) (.078) (.541) (.002) (.845) (.048) (.441) A2 0.31 0.22 0.02 0.12 -0.01 0.00 0.13 (.001) (.002) (.742) (.089) (.885) (.947) (.071) A3 0.16 0.22 0.17 0.07 -0.00 0.03 0.06 (.029) (.002) (.019) (.312) (.957) (.665) (.390) A4 0.24 0.08 0.14 0.04 0.12 -0.00 0.11 (.001) (.261) (.051) (.549) (.111) (.956) (.143) A5 0.10 -0.06 -0.08 0.29 —0.05 0.01 0.17 (.188) (.424) (.299) (.001) (.463) (.894) (.023) A6 -0.02 -0.00 0.09 0.05 0.02 0.24 0.25 (.819) (.951) (.244) (.481) (.748) (.001) (.001) A7 0.01 0.05 -0.10 -0.12 -0.20 -0.00 —0.20 (.919) (.520) (.172) (.102) (.006) (.993) (.006) Note: Factors A1 to A8 are from Side "A". Factors B1 to B8 are from Side "B". Parentheses indicate the significance level. loadings found in each factor. When the number of correla- ‘tions computed is this high, it is not unusual to find a anumber of significant but meaningless correlations. None of the correlations computed in this analysis Jreached the criterion level of .40, although many were fistatistically significant. Once again, those factor pairs lidientified as similar in previous comparisons performed ‘1? 1 COMPARISON ACROSS SAMPLES Frankfort - 8 Tawas - 9 COMPARISON ACROSS REGIONS Side "A" 8 Side "B" 7 Percent of Variance Explained: - Factor to Factor - Matrix to Matrix Within 22 for all 27 factors Within 52 across matrices Within 22 for 26 of 27 factors Within 52 across matrices --—-----------—---—-—-----‘—— Configuration of the Loadings A1 with B4 A2 & A4 with B1 A3 with B2 & B3 Complexity: - of the Variables - of the Factors None in F1 with F2 with common T1 & T6 T4 None in A1 with A2 & A4 common B4 with B1 F3 with T2 A3 with B2 & BB Communality of the Variables 4 remained in the upper 4 remained in the upper Correlation of Factor Loadings F1 with T1 & T6 F2 with T4 A3 with 32 & BB Correlationyof Factor Scores No correlations were > .40 Root Mean Square (RMS) No coefficients were < .10 No coefficients were < .10 Coefficient of Congruence (CC) F1 with T1 F2 with T4 Salient Variable Similarity Index (S-Index) F1 with T1 F2 with T4 130 Table 38 Configuration of the Highest Loading Variables for Dimension 1 -- Environmental Excitement VARIABLE F1 T1 T6 A3 B2 B3 accessible clean 0.41 secluded expensive open forested friendly peaceful pleasant middle-class- oriented sandy 0.43 0.85 0.48 fun 0.62 0.52 0.40 0.44 0.44 family-oriented 0.47 courteous enjoyable 0.74 0.56 0.38 0.50 scenic unspoiled colorful 0.42 0.63 0.61 0.60 appealing 0.46 0.42 delightful 0.48 0.52 outdoor-oriented 0.46 0.52 interesting 0.43 0.52 0.58 0.40 tourist-oriented quiet natural restful developed 131 Table 39 Configuration of the Highest Loading Variables for Dimension 2 -- Undeveloped Tranquility VARIABLE F2 T4 A2 A4 B1 accessible clean secluded 0.48 0.40 0.43 expensive . open 0.46 forested friendly peaceful 0.44 0.46 0.44 0.48 pleasant 0.51 0.42 middle-class- oriented sandy 0.48 fun family-oriented courteous enjoyable 0.40 scenic unspoiled 0.55 0.41 0.48 colorful appealing 0.48 delightful ~ 0.48 outdoor-oriented interesting tourist-oriented quiet 0.63 0.65 0.72 0.77 natural 0.45 0.49 0.50 0.53 restful 0.42 0.57 0.56 0.51 developed ' 132 Table 40 Configuration of the Highest Loading Variables for Dimension 3 —- Service Orientation VARIABLE F3 T2 A1 B4 accessible 0.43 clean secluded expensive open forested 0.65 0.41 friendly 0.40 0.50 0.58 0.53 peaceful 0.38 pleasant middle-class- oriented 0.46 0.52 sandy ‘ fun family-oriented 0.47 0.60 courteous 0.67 0.45 0.54 0.54 enjoyable scenic unspoiled colorful appealing delightful 0.39 0.44 outdoor-oriented interesting tourist-oriented quiet natural restful developed those factors were: secluded, peaceful, pleasant, unspoiled, quiet, natural, and restful —- terms which reflect a serene and relaxing setting. The dimension was named Undeveloped Tranquility. The final dimension was found in four factors from the two comparisons (Table 40). It was comprised of two varia- bles which appeared in all four factors -- friendly and courteous; and four variables which appeared in at least two 133 factors —- forested, middle-class-oriented, family-oriented, and delightful. With the exception. of forested, these variables all relate to a sense of hospitality towards tourists. This dimension was labeled Service Orientation. This chapter has presented the results of two investi- gations into the invariance of the factor structure underly- ing images of tourism destinations; one concerned with factor stability between samples for the same target region, the other with factor stability between regions for the same sample. The methods employed to assess stability in each case entailed both visual and vector comparison approaches. Three dimensions underlying respondents' images of the target regions were found to be stable in both analyses. The three dimensions were labeled: Environmental Excitement, Undeveloped Tranquility, and Service Orientation. The findings of the present comparisons are synthesized and their meaning interpreted in the following chapter. be 36 CHAPTER V DISCUSSION, CONCLUSIONS, LIMITATIONS, AND RECOMMENDATIONS Discussion and Conclusions The findings of the present study were obtained from the comparison of factor structure across samples and across regions. Before discussing these findings, however, it is important to realize that this analysis was not concerned with quantitative differences between destinations or between tourists' perceptions of those destinations. Instead, the focus was on differences in the important qualities present in the data —- the dimensionality of the data. As Stewart (1981) points out, a dimension does not indicate how much different various entities are, just as knowing that weight is an important physical attribute does not indicate how much heavier one object is than another. Quantitative differences may very well be important, but the identification of a particular dimension does not provide that information. In short, understanding the dimensional structure which underlies a phenomenon, in this case a destination's image, is a logical first step to understand- ing these quantitative differences. Thus, this analysis is best thought of as a preliminary step for future investi- gations into how visitors perceive tourism destinations. 134 135 Given this base, several of the findings from this analysis are worth noting here. First, both of the comparisons conducted —— across samples and across regions -- found the same three dimen- sions underlying respondents' images of the study regions -— Environmental Excitement, Undeveloped Tranquility, and Service Orientation. While the number of factors and the configuration of the variables comprising each dimension was not identical, there was enough convergence on these indica- tors to clearly distinguish the three dimensions. In addition, all three of these dimensions have appeared in previous investigations of regional images in other factor analytic studies. The factor termed Environment appeared in an investigation of the dimensions of tourist satisfaction with Cape Cod, Massachusetts by Pizam et a1. (1978) and is very similar to the dimension Environmental Excitement which was found in this study. The factor in that study, was made up of the following varia- bles: scenery and natural attractions, quality of attrac- tions, and environmental quality. In addition, the dimen- sion Environmental Excitement generally compares well with Craik's (1975) factor beautiful-picturesque -- each conveys a good feeling derived from the landscape. The two other stable dimensions, Undeveloped Tranquility and Service Orientation, can be found incorpor- ated in the factors which emerged in McCullough's (1977) investigation into the images of tropical destinations held 136 by experienced "long—haul" travelers and travel agents. In that study, these two dimensions were embodied in three factors which were labeled natural—sophisticated, organized— unspoilt, and service oriented-cultural. Also, Craik's (1975) factor serene—gentle is composed of many of the same variables as the dimension Undeveloped Tranquility which was found in this study. Finally, Pizam et a1. (1978) also identified the dimension Service Orientation in the factor he referred to as hospitality. In that study, this factor was composed of the following variables: willingness to aid tourists, general friendliness, general courtesy, and general hospitality. In regard to the generalizability of these three dimensions to non—coastal destinations, two of the dimen- sions found in this study appear to be similar to the factors found in Craik's (1975) study of individual varia- tions in landscape descriptions which used a non—coastal region as the target of respondents' images. This finding .provides an indication that the dimensions underlying destination images may be similar, in some respects, for both coastal and non-coastal target regions. Additional, more definitive research needs to be conducted to determine if this is the case. The identification of these three stable dimensions are of considerable importance for those involved in tourism research and the marketing of tourism destinations. For researchers involved with assessing how tourists perceive a m.‘ 81 P1 re th Dr ti ti De Ve Pr 137 given area, this findings serves to validate the contention that at least some of the dimensions involved in the factor structure of destination images remain stable across different groups of respondents and across different regions. These dimensions can be incorporated into the research instruments used to measure regional perceptions and, further, may be used to develop a general instrument which can be used for a range of target regions across a variety of respondent types. From a marketing perspective, these stable dimen- sions can be emphasized in promotional campaigns which are either general to a range of destinations or are specific to a particular destination. Thus, the Michigan Travel Bureau may include these aspects in broad promotions designed to attract visitors to the state as a whole, or may use them to promote specific destinations such as the Upper Peninsula. Finally, since these dimensions are stable across regions, they may also be used to position destinations from the perspective of the tourist. The dimensions used in a product positioning analysis from this particular applica- tion should be the more controllable dimensions of the three (for example: Environmental Excitement and Service Orienta- tion). This type of analysis could be used to compare perceptions of competing destinations, compare first-time versus repeat visitors' perceptions, or track changes in how a destination is perceived over time. In addition, a product positioning analysis can be used as a first step in 138 selecting target market segments for promotional strate- gies. For example, after completing a product positioning analysis, the results can be used to select the segment with the most favorable image of the destination in question to build upon the favorable image, or, alternatively, select those respondents with the least favorable image in hopes of improving it. In summary, these results lend themselves well to both measurement and marketing uses. One must recognize that the identification of three stable dimensions underlying destination images in this research does not exhaust the possibility that additional stable dimensions may exist. In fact, most of the studies presented in this research included at least eight dimen- sions to describe an area's attractivity (for example, see Gearing et al., 1974, and Pizam et al., 1978). The inability to identify more than three stable dimensions in this research may be explained by several reasons. The first involves the constraints imposed by the data set. The data used in this study was not collected with this particular analysis in mind. The original study was chiefly concerned with environmental descriptions of the target regions. As such, respondents' images of regional attributes (for example) were not included in the analysis. In addition, in some cases, only one variable was included among the 27 adjectives comprising the reduced data set to describe a particular aspect of a region's image. For example, the variable expensive emerged as a specific factor N: we in lIe 139 for both sides of the state in the comparison across regions. However, since only one variable was involved on this factor for both matrices, it could not be considered as stable according to the pre-established criteria. Neverthe- less, it is reasonable to expect that this dimension may be quite central to one's image of a particular tourism destination. The second reason for not finding additional invariant dimensions is that differences may well exist in how people use everyday language to evaluate tourism destinations. For instance, what may mean quaint to one person, may mean backwards to another. Further, the orthogonality imposed on the initial factor matrix by factor rotation may not be realistic considering the interrelationships that normally exist between the compo- nents comprising the totality of a given tourism destina- tion. For example, one is likely to find quality eating establishments in an area known for plush hotel accommoda- tions -— dimensions which researchers almost always consider independently, but which tourists are likely to take into account as one in their mind. Thus, the variables used to operationalize specific components' of regional images originally hypothesized as inclusive in the adjective checklist may not have been commonly understood by the majority of the sample of tourists, or may have been mixed in with other components in tourists' minds and, therefore, were not identified as stable in this analysis. p: 5C ti. th 01‘ 140 Another finding which surfaced in this analysis concerns two of the measures used to assess factor similar- ity -- the root mean square (RMS) and the correlation of factor loadings. The inability of these two measures to indicate similarity for a given pair of factors when other measures found that pair to be similar may be explained by one of three reasons. First, the factors compared may in fact, not be similar (according to these measures). Second, the criteria established for these measures may have been to strict. And third, problems with the data, rather than problems with the measures themselves may have caused the questionable results for these measures. In particular, the dichotomous nature of the original data may have caused the resulting RMS coefficients and factor scores (and thus the correlations calculated using those scores) to perform poorly. It was previously mentioned that generating factor scores from dichotomous data is not wholely consistent with the factor analytic model. Kim and Mueller (1978) point out that it is assumed, when calculating factor scores, that the original variables contain at least four different values (as opposed to the two values embodied in dichotomous data). A replication of the study using an ordinal scale (at the least) should be made to determine if this is the reason for the poor performance of these measures. In conclusion, the author believes that much was learned about the stability of the factor structure underly- ing regional images. Three stable dimensions emerged which 141 were invariant across samples and across regions. These dimensions were labeled: Environmental Excitement, Undevel- oped Tranquility, and Service Orientation. All three had been identified in previous studies involving destination images. 142 Study Limitations There were several limitations and assumptions which relate to the research design of this analysis which should be acknowledged. The present study is limited in the following way: 1. Only images of coastal destinations were considered in this study. Thus, one must question the generalizability of the three dimensions which emerged from this analysis. In addition, it is not known how these results would compare to similar research in regions with coastal destinations outside of Michigan. Information on the stability of destination images across both coastal and non-coastal settings and in settings outside of Michigan would enhance this investigation and provide tourism researchers with a better understanding of the congruence of regional images across a variety of destination types. 1 Regarding the comparison across samples, since only images of one target (coastal) region were compared it is difficult to assess the generali- zability of this particular analysis to other (coastal) regions. This information would complement the present analysis and permit a more complete investigation of the stability of regional images across samples. 143 The two areas used as target regions in this study may not have been defined well enough to allow respondents to clearly distinguish between the two (for example, both regions -- Side "A" and Side "B" -- included the coastline at the tip of the Lower Peninsula). This problem, of defining regional boundaries, is neither new for tourism research in general nor for research in Michigan in particular. Nevertheless, this may be a major limitation of the present study. The sample sizes used in both comparisons, especially the comparison across samples, were just within the minimal requirements to assure reliable results for the factor analyses. As such, it was not possible to screen those samples to assure that each consisted of respondents with the same level of famili- arity with each target region. At the least, it would have been desirable to screen out those respondents who were year-round residents of the particular site being sampled. A sampling control of this nature would have assured that only members of the target population —— tourists who were on vacation in the areas selected as sampling sites -- were included in the analysis. 144 Two assumptions underlied the present study. First, it was assumed that the information derived from the adjective checklist adequately represented respondents images of the two target regions. And second, it was assumed that the validity and reliability of the 1982 Frankfort— Tawas Study, an original data collection effort, were sufficient for the purpose of the present study. Recommendations On the basis of the results from the present study, the following are recommended to facilitate future research: 1. An application of the results of this study to assess quantitative differences in visitor images between the two study regions and between similar subsamples of tourists. Such an investigation can be used for product position- ing analysis, market segmentation analysis, the development of advertising strategies, and the .development of site choice models. An investigation into. the stability of the dimensions of coastal and nonecoastal tourism destination regions both inside and outside of Michigan. Such an undertaking would foster the development of a general research instrument ' which may then be applied to a variety of study sites. 3. 145 A replication of this study should be conducted with larger sample sizes in order to control for respondents with different levels of familiarity with the target regions in question. This would increase confidence in the results of the study. Compare images of more than one target region across samples so that this aspect of the present study can be more fully understood. Knowing that different ‘groups of respondents utilize the same underlying criteria to evalu- ate different destinations would permit the establishment of a more general research instrument to assess visitor- images of tourism destinations. A replication of the present study using matrix comparison techniques in addition to the visual and vector comparison techniques. A replication of the present study with more clearly defined regional boundaries would improve its reliability. A replication of fhe present study using a modified measurement scale (e.g. ordinal versus dichotomous) to operationalize respondent images of the target regions. It is likely that a higher level measurement scale would improve the results of the present study, especially the ‘H. 146 generation of factor scores and would facilitate the use of other statistical tools such as cluster analysis and multidimensional scaling. 8. Further research into the stability of the dimensions underlying destination images across different types of respondent groups (e.g. first time visitors and repeat visitors, visitors and non-visitors, campers and hotel/resort users, etc.) and across different types of destinations (e.g winter destinations) would greatly expand current knowledge in this area. MU. The main purpose of this study was to investigate the stability of the subjective criteria tourists use to evaluate tourism destinations. A review of the literature relative to this issue demonstrated: the usefulness of factor analysis in image research, applications of factor analysis to the study of regional images, the techniques used to compare factor structure across studies, and applications of those techniques in recreation and leisure research. It was concluded that little, if any, research has focused upon the stability of the factor structure underlying destination images. With this in mind the present study was undertaken as an exploratory investiga— tion. For the purpose of this study, two research questions were developed which focused on determining whether any of the dimensions underlying destination images remain 147 stable across different groups of respondents for the same region, and across different regions for the same group of respondents. Because of limited time and financial resources, existing cross-sectional data were used for the analysis. The data specifically used for the analysis was taken from the 1982 Frankfort-Tawas Study, a study conducted by the Department of Park and Recreation Resources, Michigan State University. The survey papulation consisted of 287 respondents who were vacationing in the Frankfort and Tawas areas of Michigan. Visitor images of two target regions —- Side "A" (the northwestern coastline of Michigan's Lower Peninsula) and Side "B" (the northeastern coastline of Michigan's Lower Peninsula) -- were assessed through the use of two adjective checklists (one for each target region). A comparison of the factor structure underlying visitor images was first conducted across samples of respon- dents -- Frankfort subsample (N - 134) and Tawas subsample (N - 125) -- for the same target region -- Side "A". The techniques used to compare factor structure in this particu- lar analysis were both visual comparison methods (the number of factors with eigenvalues greater than one, the percent of total variance explained, the configuration of factor loadings, the complexity of the factor structure, and the communalities of the variables) and vector comparison 148 methods (Pearson's correlation coefficient (r), root mean square (RMS), coefficient of congruence (CC), and the salient variable similarity index (S—Index)) using the factor loadings as the object of comparison. A similar analysis was conducted for the comparison of factor structure across target regions (Side "A" and Side "B") for the same group of respondents (N - 188). In this analysis, all the factor comparison techniques applied in the comparison across samples were used with the addition of Pearson's correlation coefficient (r) being applied to the factor scores matrix. The results of each comparison indicated that three dimensions underlying the factor structure of visitor images remained stable in both of the comparisons. The three dimensions were labeled: Environmental Excitement, Undevel— Oped Tranquility, and Service Orientation. All three were shown to be similar to factors found in prior research involving destination images. APPENDIX A 1982 Frankfort-Tawas Study: Questionnaire, Study Sites, and Demographic Profile 149 31101107... REC'xtAiiuz: AM: TOURISM IMAGE. 51;... m‘ Park and Recreation Resources - hichigan State University We are interested in what you think about Michican. It is not important that you have been to all of the regions in the state to answer the questions. Your answers will be held confidential. . accessible . clean acrowded __:__aecluded . drab - fiat ' expensive I om ' _forested . friendly __-_.unusual __'_peaceful _-__o1easant o niddle class oriented ' sandy ° unl? ° monotonous - clear - hostile . fun l 9 - family oriented 'spirited . accepting __ ._courteous o gracious . enjoyable ' tacky. ' spectacular . primitive :5: quaint .___-_upper class oriented __ Lalivo ”Lapwing ‘ __:_bright .__;.corlaercial oriented _o_delithtfu1 Thank you for your help and cooperation! Please use the map at left to complete the followinc question. 1. The following is a list of adjectives. Please read them quickly and check each one you would consider descriptive of coastal area A shown on the map at left. - exciting . festive __-._outdoor oriented - horrible . out-of-the-way __:_1ifeless o nondescript ' noisy ' interesting __0_tourist oriented a quiet _'_ natural ° restful _;_developed __0_tasteless 0 classy . heavy traffic __..b“‘.' If you feel you cannot fill out the above question. please check why below. Haven't been there. “Haven't been there, but I'm willing to give my impressions. (Please complete this pace if you check this response) Not faniliar enough to fill this out. Other. please specify. 150 Please use the map at left to complete the following question. 2. The following is a list of adjectives. Please read them quickly and check each one you would consider descriptive of coastal area 8 shown on the map at left. accessible family oriented exciting _ ___cloan ____spirited festive crowded accepting __outdoor oriented ____secluded courteous horrible drab ‘____gracious ___put-of-the-way flat enjoyable lifeless expensive tacky nondescript. open spectacular noisy forested primitive interesting friendly remote‘ tourist oriented unusual scenic quiet peaceful unspoiled natural pleasant colorful restful middle class oriented quaint develped sandy upper class oriented tasteless ugly alive classy monotonous appealing heavy traffic clear bright busy hostile commercial oriented fun delightful If you feel you cannot fill out the above question, please check why below. haven't been there. haven't been there. but I'm willing to give my impressions. (Please complete this page if you check this response) Not familiar enough to fill this out. Other. please specify. 151 3. The following is a list of activities. characteristics and facilities that you might find while in the two areas of Michigan shown on the map at left. Thinking of these two coastal areas. please check which of the two you feel would best provide for each of the following: Coastal Coastal Area A Area 3 Tourist attractions Peace and quiet family entertainment Expensive entertainment Vacations historical interest Places to eat Places to shop Lake access Bench activities Swismung Sailing Boating Picnicking Charter fishing Fishing Hunting Photography Natural areas Observing wildlife' Hiking Camping MCanoeing S. 152 Please use the map at left to answer the following questions. Please rate coastal area A on how well it provides for recreation and/or tourism opportunities. (Please circle one) Poor Excellent 1 2 3 i S 6 7 Please rate coastal area I on how well it provides for recreation and/or tourism opportunities. (Please circle one) Egg; Excellent 1 2 3 6 5 6 7 Please rate the State of Michigan. compared to other Great Lakes states. on how well it provides for recreation and/or tourism opportunities. (Please circle one) ‘ Poor Excellent 1 2 3 e 5 6 7 Are you a resident of the area you are in now? yes no If ves. how many years have you lived in the state? years lf no, what is the purpose of your visit to this area? (Circle one) to visit friends or relatives business and pleasure business pleasure vacation other. please specify Bow familiar are you with coastal area A compared to the other coastal areas in Michigan? extremely very somewhat not very not at all familiar familiar familiar familiar familiar 153 9. how familiar are you with coastal area 3 compared to the other coastal areas in Hichigan? extremely very somewhat not very not at all familiar familiar familiar familiar familiar 10. Sex: male female ll. Age: ________;years 12. lducatioo: Please circle the number that represents the highest level ccnpleted. l 2 3 t 5 6 7 8 9 10 ll 12 l 2 3 6 5 6 7+ Grade School High School College ll. Annual family Income (before taxes) ___m« “.999 __ swoon-319.999 ___sao,ooo-s:u.999 _s§.ooo-99.999 _szo.ooo-sza,999 _sas.ooo.51.9.999 __uo,ooo-m.999 _szs.ooo-829.999 __sso.ooo and over 1‘. when thinking about the area you are in right now, what one thing stands out most in your mind? 154 SAMPLE SITES: SUMMARY Location: East Tawas Dock Date: August 16 Time: 1&00-1600 Weather: Partly cloudy. windy Conditions: Many families walking along dock. Survey site was at only pt. of entrance on and off dock. Passed out surveys while stationary at this Pt. Location: Beach adjacent and north of East Tawas Dock Date: August 16 Time: 1615-1700 Weather: Partly cloudy. windy Conditions: People were lying or sitting on beach. Few swimming. Made one pass up beach and asked everyone on beach (12 and older) to fill out ‘ questionnaire. Location: Campground at Tawas State Park Date: August 17 Times-lOlS-lZIS Weather: Partly cloudy Conditions: Flipped a coin and started an east side of campground and sampled sites sequentially. Asked campers at all occupied sites where campers were visible outside of their campers or tents. Location: Tawas State Park Day-Use area Date: August 17 Time: 1215-1425 ~ Weather: Partly cloudy with darker clouds on west horizon Confitions: Sample was from the north end of the beach southward to the point where the beach narrowed down from approximately #0 yards to 15 yards. Made one pass (north to south) and asked each group onobeach during that time interval. Location: Tawas City Resorts (south end of city) Datt: Sugust 17 Time: 1530-1700 Weather: Partly cloudy. getting darker Conditions: Requested permisi n at four resorts and received permission to solicit people outside of their rooms. Bulk of those questionnaires can from one resort. Other resorts either had no people present or managers were not available to request permission. Surveys all can from 3 or 4 loosely organized groups. 155 SAMPLE SITES: SUMMARY Location: Park right next to Pollice Station in Frankfort Date: August 18 line: 1400-l600 Weather: Partly cloudy . Conditions: Foot traffic was not very heavy with most stores being across the street. Was stationary for this sample. Location: Frankfort beach (west end of town) Date: August 18 Time: 1615-1815 Weather: Partly cloudy Conditions: Started on beach below parking lot and worked up beach about a quarter mile to point where small cliffs meet the beach. Sampled peeple on way to and from cliffs as there were low numbers of peeple and the interviewer was able to pick up newcomers to beach in this manner. Location: Mineral Park (two blocks east of Police Park) Date: August 19: Time: 1000-1200 Weather: Overcast. threatening to rain Conditions: Park was about five acres in size with a covered picnic area and marina. Data was collected along the marina and at picnic area. Just before noon it started raining and last few peeple.filled out surveys under cover of the picnic area. Location: Lake Park (Honor: about 12 miles north of Frankfort) Date: August 19 Time: 1300-1600 Weather: Very overcast and windy. No rain however. Conditions: Collecd most questionnaires as people landed their canoes at Canoe Landing on Platte River (canoes were coming from two canoe liveries about three miles upriver). Collected a few surveys from people parking and then walking to the beach. - Demographic Profile by Subgroup SEX OF RESPONDENT Male Female Incomplete Date Total AGE OF RESPONDENT Under 24 25-34 35-49 50-64 65+ Incomplete Date Total Mean ANNUAL FAMILY INCOME Under $4,999 $5,000 to $9,999 $10,000 to $14,999 $15,000 to $19,999 $20,000 to $24,999 $25,000 to $29,999 $30,000 to $34,999 $35,000 to $49,999 $50,000 and over Incomplete Date Total Median LEVEL (YEARS) OF EDUCATION Grade School (11 or less) High School (12) College (13-16) Graduate School (17+) Incomplete Date Total Mean 156 Table 81 Frankfort Respondents N 2 62 45.62 67 49.32 7 5.12 136 1002 N 2 21 15.42 30 22.12 47 34.62 21 15.42 9 6.62 8 5.92 136 1002 39.T N 2 6 4.42 4 2.92 7 5.12 8 5.92 19 14.02 14 10.32 9 6.62 33 24.32 18 13.22 18 13.22 136 1002 $30-$34,999 N 2 7 5.12 29 21.32 55 40.42 39 28.72 6 4.42 136 1002 15.0 Tawas City Respondents N 2 73 48.32 71 47.02 7 4.62 151 1002 N 2 38 25.22 51 33.82 35 23.22 16 10.62 2 1.32 9 6.02 151 1002 32.5 N 2 9 6.02 8 5.32 7 4.62 11 7.32 20 13.22 29 19.22 13 8.62 19 12.62 9 6.02 26 17.22 151 1002 $25—$29,999 N 2 21 13.92 61 40.42 41 27.22 21 13.92 7 4.62 151 1002 13.3 Combined Respondents N 2 135 47.02 138 48.12 14 4.92 287 1002 N 2 59 20.62 81 28.22 82 28.62 37 12.92 11 3.82 17 5.92 287 1002 35.9 N 2 15 5.22 12 4.22 14 4.92 19 6.62 39 13.62 43 15.02 22 7.72 52 18.12 27 9.42 44 15.32 287 1002 $25-$29,999 N 2 28 9.82 90 31.42 96 33.42 60 20.92 13 4.52 287 1002 14.1 APPENDIX B Listing of FACCOMP Data Analysis Program 157 .739: 227*— 2:—F<.: :p.._.: ZC 0,: Zn... 2: ........:...z z: .(a .CCu :35 $3.42— ..2:_.13 .n.33~ ml:_::;; tCIZZLL mzh .03:z\mm.&C—;c} ¢:n :zc nztdeU\$ZC—U(u cm .vzdu—ni3 a..;= tZt—KCt C u: safpiatzu mug—Zth :CFQCm 4:5 .zkmfla an tcxocx; USP ..2_u——;3 . r.um_._§:u.,.:::2 2:4: :ZC wzo~h¢——t—J.cz Gus—3 I. .zzuhflba . .a..fl:9~ no .23.. Zuhhnzl mC) tcanZL w:— ..xu.=:lmv x92.— : .VZJUp—zz E A A. - .A . .A I f :.»-....<;::m ”EC—.2) FZb—dcm urn :Zc :qu hazyzczsu as. Z.u_U~Luuc.u H2... .czap—c: : .anmtg. mac-Em zcut #001 5:, to um: .ra mu—aabm maomuc mxcpucu wan-(.utou .VZAu—nxa “f:— ._:.zcwmu: .raaozmmé .5 .:.,.c._ .22 20¢”..(fi .n. mast... .C- zuhhuvu: .535»... < .szUpnti .a..a:ouo .uaau.cca-.mzosA-z:-| At. A A , ...A A AA ,- - . -. - .cseux . x ac)- .cacwc . me: x. aausupza . . u.a.c, .aoasoz_¢pms . naauxopoca .....ex.zpms .oapnus.a.. . . , . .ionscz_cpm. “can one: . .nzmom_z_ . nu:ox-cz ..n¢uoupmmAa.nmxc:soowox. .aueupza . > .x - .mxuou_z_ . .mmaoznoz .mzueupz_ . .mzczsou a: .xoasoz_¢pms . .o_-ao»o¢¢¢< .. ss»-m:x .nac ac.non... .ccn....»ca¢< a muses axons—z. . sea:— - ac: .mzu A A A - .532: a piaurxéAF-.- 1 A.e A: - -IA . , .o.cam a m excuse.. ma»_uxusz.-o - A.,.. -A ,- : A ., .. ,iszu; .vseux . u>::.m . , .. a .. “edema..-..»=:»m - A.- _ A. A- . exouux.. mac . ma»— :pzkpao. :52— .nEOuuC... £192.; 158 :2: so... a.-. e .: 2 .eezu:a.. . 33232.. 2.3.... 2: 2:... .2. 3.3:...3223 5.: as: sea as .mzu - : numaee.ua zero-s_aes ..... 3.3:... 3:. as. as: :55 See: me: a. 3:5... zero 52: as: 2...: .. 23:23:13.5. axe 3a. A 33.223235: c: :2. A. 2.....zeze_.a.._c.: are :3 H 3.2ucezaza...¢o: a. .23.. - . !- .2 e o. 7...-.. can Ended... eexu ease: 2226552233.... e A . . Essen... 32-93 22232.32? 2.. on» x22: .3 use: $39522 .euaue . - .ucezuda... uaodcez 8.5. see 3.. .52: - 3.3 .35.... as: sea 2.22.323 . 3:35... 13.328 . .23. so. use: .2: of... £2 3.3:...2323 tun—um I!» a v C bu - 35 .35... 3: ae- 2ucezuzef . fez-33s {savaged a .95. 2: was: .25 use: £2 2.332.532: : I :_soa as: e . e a. :utCZudua. . asceuopeocu 3:2: 02. see: a...» m::>c—..zdup~§ ZN: wanuuacn Lu-.- . c . - - 33:... .ce..u-a3es!l-i - . rest... 33-.....- 22» seeds—5e: mo: .2. .35.... 3:8 .225 5.: 22. , 2.3523235. see 3. A .nasutexw2aeai... . : co 2; .. 2:223:58. axe 3e \, 33:32:29. 2 :3! me e a. a... a 8.. .322: 35-3.2... 222.25.32.23... e , 355.... 22.5... .Gcezudaszseue 2.. as: x22: no are: $39.53.. pecans . . . . . :. . : ..¢=—uea.xaecu¢ A- 2. . Eupueeezu use. on» utter i... .: ensued Scale—:3 .32e3e:.eieue 2. . 22.....er are. are :22: cc. .2 Speed eu—ELutes ‘Izdosu ’3 pence. -. . - -2332! on a 3 .2... exact .ateue 2.:- we.:zou op quota” amuse :- .5225 . . ,- -2. Leann—.5 2..an rice 2.: e: on 552:8 55:: e E Site use euunepec a: .5325. 21323.... x. 32:; a: aeoeue 323.: 9:5 Epuea em escapes—econ. U: 2.3.3:! 2.:e ace escape 5o 2.. a: «panacea 25$. :ueu 33.2.4 a: sea on— axe , tease—:3 218 u: use use .9... u:— coa use Ems—See to 33 use music... seems... i: Lean—:3 2.25.23.35.33 2. 2.22:! 2.25: 555.. e we :95 mam—ueaeru 2.3.3 .3352: 2. are xaepxou Eases see nee-us}: :33 .2... sense. oz us 3.... use: use .2222: 2. a: 2.: use... flier cues to area as: a: m. .533 u:— ictea sues Lad—2e. X ‘0 159 . 2.3.. a a: .nz..2.au .oz ...:2 ..2-;e> .m2..:e.2 - ‘qolw A 2.o.2 on .nz...0u .oz u..21ru I a A .o... “222:.ouoo2 .o... amaze-a: "do...“ c ..n..222:.ou-oz.... . .22.. 2°. m22:.c. .o oz ..n..m2o2noz...a . >azpu .o. 2:92 .o o2..:.a...a- . A .n... .u22:.ou :2 . - --.:-..su|.--- A-sA--.. .- A.-.-.. u-e...AA ;.u:o2-o2:u. . A. A A A A - . , A A . . .-: .azu J,. ...2..242euz . A - A- . .. + e... e ,. - .2 2e)... .»a=.a..n.e.mumeu . .azu - A-A;--A. - -. . ...x-2e:..2..aeua . - t A . -- n - :. - A. :-A A A A - - ; .. . a... a . . .x-2e>.-. .»a=.u.nn.e.mumeu A :, A A A A . A 2.2.2A --- Ax - A AAA-A A. --: . .A - A : .A A A on ..2..2.au .a2 u..2: . .22-2e: ..2..aeux .. . . lulllulcu. .l- I! . II.- .. i :II I , I 3.! ~. vii-Inl’- cs. 1....“qu I - . . A A A , , A - 1 2.2.. a . . on ..2...au .oz u..22 . , A..-A-AI!-. -A.- Aft-12.-.?! .aztasu-oxA.Ao.-.A.-A.m....¢-o2.:-i-- -, . - A A A, A A . A. A-, A- -- ..... e , A A .. . .A u- A 2.awa . A A :22. 22:2 .. v - .o..e...m ~22; 2.2» ..o... 2e2u 2.. 22 9.2... «exu 2.. .. A-.A- - - A A A - . . .uzoa ..»2: _ .- .3,:- -. -.A i if A A3.... A AA -- -, -.- .; A .-;l - A. A. A A : - -:-- - .o...m2¢€n2. . - .o... mm;...2. ..m.e..n. .222 2.2. ...... 2&2. 2.. 22 ..2..- 2&2. 2.. .. - AA . ...... xezu-2.. 22 ..¢.. 2e2u-2.. 2: ..e.. 2e2u 2.. 2c ..e... 222g 2.. ...2: » ..2e:u-2..2.=e.2 ....2... .e.>.2.uz >2.u2 22 ..e... .eu>.2»u2 .2oae..u..23 pew..2 2. . 22...... 2.2 ~w.u.......2......5 o 23.-9. A 2.2. .o... 222...2.. 22 .o... m2:..a2.. .. - - A ..n2...umu¢ » ...2...umu2 .u:2.... 2e...m2. .4e..2u=cum... u2oe.m2. o .222»... .ez»..2. ..e..mu:c.m.-. u;c2..2. ..c:. axe: .n2.. 2¢.mue . A ...2a-u2e2 ..2.. 22.nue I’. . U.‘ ' A A - A A: - A-. A A , ;A A .utezu.mw.n. capcutem . .m... sea .22 ...2: 1|)- . , u .22. . .uS¢—.I. uduu ac: ..0£oco..b.. . .6312... 20.90.25 . Juana ooAu 0‘2. 0...: :ec.o~: o.~«..>.:—..z..u-¢3 . 35:-.- . - - - . A) I -. .. - - - Snow; 0 8*: a .. c .3 - .22. b .u:2..-. e... 2e. .AUtCZaA: .. . . ~.O...I- :2. as... 3.59:... O .9::~ 8a :9: On -,3 I07: O-nb u: ~9)¢-. vIdUbns 160 “:3; J n.x-;<>~z_;cmx . ..au>up;.:._;.-uu.. .4 uu ...:;.:c.c_z...: . ..n>;:»m-::w . _»;:»u-tzm._z:m \.n;.c_::u.‘_.na.c..z.-uu J 22923555...- c..m..z_m:x: . «:zu *na.¢.:=a + panach....nn.c.c=w ..o."n_u__-:4=:.. ..nuu...4 u «aK.. p. :.a:.._:g.=o.zdup_xa . . .n»a:pm.na.uumuscu . _»a:_m.nc.uumuma=pu-::a..~ .»;:pm-::m a ” .oo.o... ccmnnac: . ““. . . . . . . . u z_oua a: «was: oz an «.n. u to; m ‘ .‘.. ‘.:, _ ..;‘ x . . . - _. .oo.o.u. «banpuucnm “ “oo.o.u. u»a:pm cam . . .o.o.-..na.c.::m . - ‘ , - - n ‘ z_ouo . o: nuznaaou oz up .... a xou I x~aux -il-‘. “‘l ,‘- - ‘ -. . -,- _ II‘ .. ‘ . _ ; o: «szSJOu 02 an «cu» C 195 _ . ,- - 4‘ -- :- .. - -. - --.. :-.3.:: . - .nzu z- .azu ..x-¢<>.:4:a:_a gob mztaaou go a: ..n,nmaaz-oz...a n >aapa no. use: uo oz..:‘ua..: , ,-. .. - . ; I.-.... .u.. ;.a...,mm2::4auuoz .,u‘ a. . ‘ .. u c... nmzoc oz \ .azu ‘ . - , . . .. _ -...nz~.zda¢ux .. . <... c v .x-¢c:.u. «paapm.nn..zuhz_:a.u»~x3 z_cum c: nmzz:4au-az a" ..u. c ac: «on c c «as!» U . 1 . ..cuh2_¢m.zauh~u: a n.. o...»chu.¢upz_¢a.u»~¢aV. .- . .. z—cmn v.2 on umztaaau-oz a» ..-. c me: an; - - ; , . . . . . ., ‘ , ‘-, . . - .-- .‘ ., - n..a. u - A ‘ l . ‘.-; ..¢upz_ca.zaup_za ..‘ m....z:~ucu.. n‘._"¢o_u: ..n.ouw.na.c.xuaz_-a.. .-,l| I'.II. . :. ..o.»z:ou»z.nn.<.xu:z_ua.. ..¢upz_¢m.umo49 ..¢u.x.¢m.ue¢¢ .azu -..¢u»2_¢a.zauh_¢a .. . u..” u ..¢w»z.¢m.:.up.¢a .a2w ..¢upz~¢L.u~_¢a 2.9»: o: «mzzadcu-oz up «.nu . mom ..n~u..~Uan¢_z¢>.xupz_za.up_¢a z.aua on nmzzndau a2 a» ..1” c cog “0.4“ U .Azu»z~¢a.xau»~¢3 ...n.._“xo_ucu.. , n {. ;. ”sopucu.. ..z».z_z;.z.u__za ..¢..z_x4.z.u._xa ..¢u_z_¢¢.z.up_a: ..¢w_z_¢m.u~_xawx .....p¢4..¢u~z—¢a.zo_mmc -..¢u~z~¢¢.umc4u.l ..¢u»z_xg.uo