INFORMATION TO USERS This material was produced from a microfilm copy of the original document. While the most advanced technological means to photograph and reproduce this document have been used, the quality is heavily dependent upon the quality of the original submitted. The following explanation of techniques is provided to help you understand markings or patterns which may appear on this reproduction. 1. The sign or "target" for pages apparently lacking from the document photographed is "Missing Page(s)". If it was possible to obtain the missing page(s) or section, they are spliced into the film along with adjacent pages. This may have necessitated cutting thru an image and duplicating adjacent pages to insure you complete continuity. 2. When an image on the film is obliterated with a large round black mark, it is an indication that the photographer suspected that the copy may have moved during exposure and thus cause a blurred image. You will find a good image of the page in the adjacent frame. 3. When a map, drawing or chart, etc., was part of the material being photographed the photographer followed a definite method in "sectioning" the material. It is customary to begin photoing at the upper left hand corner of a large sheet and to continue photoing from left to right in equal sections with a small overlap. If necessary, sectioning is continued again — beginning below the first row and continuing on until complete. 4. The majority of users indicate that the textual content is of greatest value, however, a somewhat higher quality reproduction could be made from "photographs" if essential to the understanding of the dissertation. Silver prints of "photographs" may be ordered at additional charge by writing the Order Department, giving the catalog number, title, author and specific pages you wish reproduced. 5. PLEASE NOTE: received. Some pages may have indistinct print. Filmed as Xerox University Microfilms 300 North Zeeb Road Ann Arbor, Michigan 48106 LANIER, Louis Legrand, 1926* AN EXPLORATORY STUDY OF USE PATTERNS AND USER CHARACTERISTICS OF MICHIGAN SNOWMOBILE OWNERS. Michigan State University, Ph.D., 1975 Recreation Xerox University Microfilms, Ann A rbor, M ich ig a n 4 8 1 06 AN EXPLORATORY STUDY OF USE PATTERNS AND USER CHARACTERISTICS OF MICHIGAN SNOWMOBILE OWNERS By Louis Legrand Lanier A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Resource Development 1974 ABSTRACT AN EXPLORATORY STUDY OF USE PATTERNS AND USER CHARACTERISTICS OF MICHIGAN SNOWMOBILE OWNERS By Louis Legrand Lanier The tremendous increase in number of snowmobiles, from 196 2 to 1970, was cause for some concern to planners and managers of public land in Michigan. The problem selected for investigation was the question of whether or not significant differences in use patterns and socio­ economic characteristics exist between various groups of snowmobilers and, if so, of what significance are these differences in managing resources for snowmobiling. Since Michigan is composed of three rather distinct geographic regions (Upper Peninsula, northern Lower Peninsula, and southern Lower Peninsula), it was of particular interest to determine if there are differences between the snow­ mobilers in these three regions, and if so, in what way are they different. An additional aspect of the problem ivas that of examining methods used to collect data. A survey of registered snowmobile owners was con­ ducted in June and July of 1970 in cooperation with the Louis Legrand Lanier Michigan Department of Natural Resources. A four-page self-administered questionnaire was designed requesting information about: the household; (1) socio-economic characteristics of (2) opinions of respondents concerning regulations governing snowmobile activity; (3) patterns of snowmobile use in terms of number of snowraobiling days during the 1969-70 season, counties used, ownership of land used, kinds of snowmobile activities, other recre­ ational activities associated with snowmobiling, and length and duration of overnight trips; and (4) information related to the snowmobiles owned by the household. A random sample of 5,133 registered snowmobile owners, stratified by region, was drawn from the state­ wide listing of 127,769 registrations. The questionnaires were mailed on May 26, 1970; a reminder card followed by a second questionnaire were sent to nonrespondents on June 16 and 26 respectively. By the cut-off date, July 14, 3,527 usable responses were received: percent. a response of 70 Sub-samples of both respondents and non­ respondents were interviewed by telephone to determine if differences existed between these two groups. No sta­ tistically significant differences were found. Contingency tables were developed for each variable across regions. F test were significant If the analysis of variance and (.05 level), then Scheffe post- hoc comparisons were made to determine which pair or pairs of means contributed to this significant difference. In Louis Legrand Lanier the case of qualitative variables, an overall chi-square was calculated; if it was significant then the chi-square was partitioned to determine where significant differences la y . The Automatic Interaction Detector (AID) technique was employed to determine the presence of interaction effects between "predictor" variables and to determine the relative importance of these predictor variables in accounting for the variation in the dependent variable (number of snowmobiling days).^ Small regional differences were observed in the age and educational level of the head of snowmobile-owning households and in the number of children, 18 years old and under, in the household. However, substantial dif­ ferences existed between regions in incomes: income. the more urban the combined household the region the higher the average There were differences from one region to another in the occupation of the head of the household. Attitudes of snowmobile owners toward regulations governing snowmobiling activity varied little between regions, except for attitudes concerning the use of town and village streets for snowmobiling. In general, snow­ mobile owners favored stricter regulations and greater enforcement of them. Snowmobile owners in the southern Lower Peninsula went snowmobiling fewer days than those regions. The more rural in the other two the region the more snowmobile Louis Legrand Lanier owners tended to go snowmobiling in their county of resi­ dence. Snowmobilers in the southern Lower Peninsula spent a large portion of time outside of that region while snow­ mobilers from the rest of the state confined snowmobiling activity almost exclusively to their own region. In general, respondents spent a greater proportion of their time snowmobiling on privately owned land than on pub­ licly owned land. It was, also, noted that snowmobiling was more than just a sport by itself, but, was frequently associated with a number of other activities. The results of this study lead to a number of recommendations. The main ones were that: (1) spacial units used for collecting and analyzing data should be more homogeneous, be developed, (2) a predictive model of snowmobile use (3) high priority be given to investigating user from the urban areas in the Lower Peninsula, and (4) in order to understand the behavior of the resource user it is necessary to examine their life style, rather than just their behavior in one activity. "Predictor" is used in this phase of the analysis in terms of explaining the amount of variation in the dependent variable; not in terms of developing a functional equation. ACKNOWLEDGMENTS The completion of this study was made possible by the assistance and cooperation of many people. I am par­ ticularly indebted to Mr. Norman F. Smith, Office of Planning S ervices, Michigan Department of Natural Resources; Mr. Paul Rearich, Parks Division, Michigan Department of Natural Resources; Professor Louis F. Twardzik, Department of Park and Recreation Resources, Michigan State University; Dr. Michael Chubb, formerly the Director of the Recreation Research and Planning Unit, Michigan State University, for making this study finan­ cially possible. Dr. Philip M. Marcus, Department of Sociology, Michigan State University, not only offered counsel and encouragement but arranged for the Urban Survey Research Unit to carry out the questionnaire coding and the transfer of data to computer tape. Mr. William H. Colburn, Office of Planning Services, Michigan Department of Natural Resources, offered assistance throughout, in particular, with the printing and distribution of the questionnaires. I am grateful to the untiring assistance provided by the staff of the Recreation Research and Planning Unit, Michigan State University, in carrying out many of the routine, but necessary, tasks connected with con­ ducting the survey. To Dr. Daniel E. Chappelle and Dr. Milton H. Steinmueller, Department of Resource Development, Michigan State University, I am appreciative of the constructive criticism and encouragement they provided. Dr. Dennis Gilliland, Department of Statistics and Probability, Michigan State University, gave generously of his time and advice. I am most grateful to J. Paul Johnston, Department of Political Science, University of Alberta, for his invaluable assistance with the last phase of the analysis. Finally, I am most indebted to Dr. Michael Chubb, formerly with the Department of Resource Development, Michigan State University, for his unfailing support, constructive criticism, and guidance throughout the study. TABLE OF CONTENTS Chapter I. II. Page STATEMENT OF PROBLEM ....................... 1 Introduction........................... The P r o b l e m ........................... Objectives ........................... Significance of the Study . . . . D e f i n i t i o n s ........................... Limitations and Assumptions. . . . O u t l i n e .............................. 1 5 7 7 11 13 14 METHODS AND P R O C E D U R E S .................... 15 Data Collection........................... 15 T e c h n i q u e s ........................... Self-Administered Questionnaire . . Sample Design ....................... Distribution andReturns............... Data C o d i n g ........................... N o n r e s p o n s e ........................... Snow Depth . Methods of A n a l y s i s III. 15 20 27 33 40 42 43 45 .............................. Phase I Phase I I .............................. 45 47 Analysis problems ................. Analysis procedures ............. 47 49 RESULTS AND A N A L Y S I S ....................... Phase I 54 ................................. Socio-economic Characteristics. . 55 Age c h a r a c t e r i s t i c s ............. Education characteristics . . . Occupational characteristics . . Income characteristics . . . . 55 57 59 63 iv . 54 Chapter Page Number of c h i l d r e n ................. Membership in snowmobile clubs . Snowmobiling groups preferred . Ownership Characteristics. . . . . 71 Snowmobile horsepower ............. Years of s n o w m o b i l i n g ............. Number of snowmobiles per house­ hold .............................. Other equipment owned by house­ h o l d s .............................. Opinions About Selected Regulations 71 71 74 74 . 78 Present regulations................. Enforcement of regulations . . . Operation of snowmobiles by those 14 years and u n d e r ............. Regulation of the noise level . . Regulation of snowmobiling near ice-fishermen .................... Allowing snowmobiling on public thoroughfares .................... 78 78 Patterns of Snowmobile Use IV. 66 68 68 . . . . 80 80 80 84 88 Amount of snowmobile use . . . . Ownership of land used for snow­ mobiling.......................... Kinds of snowmobile activity. . . Activities associated with snow­ mobiling.......................... Overnight snowmobile trips . . . 89 102 106 Expanded Snowmobile Use Days. . . . N o n r e s p o n s e .......................... 109 119 91 97 Phase I I ................................. 120 Region I .............................. Region I I .............................. Region I I I .............................. Summary................................. 123 128 133 138 CONCLUSIONS AND RECOMMENDATIONS............. 14 3 Context of the Study v . . . . . . . 143 Page Recommendation 144 Study M e t h o d s .........................145 Response r a t e ..................... 145 Attitude scales..................... 14 5 Measurement of t i m e ............... 146 Recommendation ....................... Estimation of socio-economic v a r i a b l e s ......................... 14 7 Recommendation ....................... Detection of interaction effects . . Recommendation ....................... 14 6 147 147 149 User C h a r a c t e r i s t i c s .................. 149 Age level of head of household. . . Educational level of head of house­ h o l d ............................... 150 Income of h o u s e h o l d ............... 150 Occupation of head of household . . Size of h o u s e h o l d .................. 152 149 151 Patterns of U s e ......................... 152 Amount of snowmobile u s e ........... 152 Classes of land u s e d ............... 153 Snowmobiling and associated activi­ ties................................154 M o b i l i t y ............................ 156 Recommendation . . 157 Implications for Planning and Manage­ m e n t ................................... 157 Need to examine changing behavior. . Recommendation ....................... Need to forecast behavior . . . . Recommendation ....................... 157 159 159 159 Application of Recreation Research G e n e r a l l y ............................ 160 Life style of users . . . . . . Recommendation ....................... Underlying values . vi 160 161 161 Page APPENDICES Appendix A. QUESTIONNAIRE LETTER OF TRANSMITTAL AND MAP . B. REMINDER C A R D ...................................... 165 C. REVISED LETTER OF TRANSMITTAL..................... 166 D. TELEPHONE INTERVIEW SCHEDULES FOR NON­ RESPONDENTS ..................................... 167 E. SNOWFALL INDEX AND AVERAGE INCHES OF S N O W F A L L .........................................171 F. CLASSIFICATION OF VARIABLES G. WEIGHTS USED IN EXPANSION OF DATA . . . . H. MICHIGAN'S PUBLIC L ANDS............................176 I. AID A L G O R I T H M ......................................177 J. WEIGHTS USED IN AID P R O G R A M ..................... 183 K. PROPORTION OF VARIATION IN NUMBER OF SNOW­ MOBILING DAYS EXPLAINED BY PREDICTOR VARIABLES.........................................184 SELECTED BIBLIOGRAPHY ................. 163 173 175 18 5 vii LIST OF TABLES Table Page 1. SAMPLING STATISTICS ........................... 34 2. DISTRIBUTION OF AND RESPONSE TO MAILED .............................. QUESTIONNAIRE 39 AGE OF THE HEAD OF SNOWMOBILE-OWNING HOUSE­ HOLDS ........................................ 56 EDUCATION LEVEL OF HEAD OF SNOWMOBILEOWNING H O U S E H O L D S ........................... 58 OCCUPATIONS OF HEAD OF SNOWMOBILE-OWNING H O U S E H O L D S ................................. 60 COMBINED GROSS INCOME OF SNOWMOBILE-OWNING H O U S E H O L D S ................................. 64 NUMBER OF CHILDREN 18 AND UNDER IN EACH SNOWMOBILE-OWNING HOUSEHOLD................. 67 SNOWMOBILE CLUB MEMBERSHIP BY SNOWMOBILEOWNING H O U S E H O L D S ........................... 69 TYPE OF SNOWMOBILING GROUPS PREFERRED . . . 70 10. SNOWMOBILE HORSEPOWER 72 11. YEAR OF FIRST SNOWMOBILE PURCHASE 12. NUMBER OF SNOWMOBILES PER HOUSEHOLD. 13. OWNERSHIP OF OTHER VEHICLES BY SNOWMOBILEOWNING H O U S E H O L D S ........................... 76 SHOULD THE PRESENT SNOWMOBILING REGULATIONS BE MORE S T R I C T .............................. 79 SHOULD ENFORCEMENT OF SNOWMOBILING REGU­ LATIONS BE MORE S T R I C T .................... 81 3. 4. 5. 6. 7. 8. 9. 14. 15. ....................... viii . . . . . . . 73 75 Table 16. Page SHOULD CHILDREN 14 YEARS OF AGE AND UNDER OPERATE SNOWMOBILES ........................... 17. SHOULD SNOWMOBILE NOISE LEVELS BE REGULATED. 18. SHOULD SNOWMOBILING NEAR PERSONS ICE-FISHING BE R E G U L A T E D ................................. 82 . 85 19. SHOULD SNOWMOBILES BE ALLOWED TO 20. TOTAL AMOUNT OF SNOWMOBILE USE PER SNOWMOBILE OWNER DURING 1969-70 SEASON ................. 90 AMOUNT OF SNOWMOBILING ON VARIOUS CLASSES OF LAND PER SNOWMOBILE O W N E R .................... 92 21. DRIVE ON A: 83 KINDS OF SNOWMOBILE ACTIVITY PER 23. ACTIVITIES UNDERTAKEN PER HOUSEHOLD IN ASSOCI­ ATION WITH SNOWMOBILING....................... 10 3 24. OVERNIGHT SNOWMOBILE T R I P S ........................ 107 25. EXPANDED SNOWMOBILE USE DAYS - IN THOUSANDS OF DAYS ........................................ 110 26. PREDICTOR VARIABLES AND CLASSES USED IN THE AID A N A L Y S I S ..................................... 121 27. REGION I: PROPORTION OF VARIATION IN EACH GROUP EXPLAINED BY EACH PREDICTOR VARIABLE............................................ 126 28. REGION I: CHARACTERISTICS OF THE FINAL GROUPS CREATED IN THE AID A N A L Y S I S .................... 129 29. REGION II : PROPORTION OF VARIATION IN EACH GROUP EXPLAINED BY EACH PREDICTOR . . . . 131 REGION II: CHARACTERISTICS OF THE FINAL GROUPS CREATED IN THE AID ANALYSIS . . . . 134 31. 32. . 86 22. 30. HOUSEHOLD . . 98 REGION III: PROPORTION OF VARIATION IN EACH GROUP EXPLAINED BY EACH PREDICTOR . . . . 137 REGION III: CHARACTERISTICS OF THE FINAL GROUPS CREATED IN THE AID ANALYSIS . . . 139 ix . LIST OF FIGURES Figure Page 1. The snowmobile questionnaire ................. 24 2. Regions and selected counties used in the sample stratification and analysis . . . 30 3. 4. 5. 6. 7. Net gain or loss, in thousands of snowmobile d a y s ........................................ 116 Percentage of gain or loss in snowmobile d a y s ........................................ 117 AID "results tree": an analysis of the number of snowmobiling days in Region I . 124 AID "results tree": an analysis of the number of snowmobiling days in Region II. . . . 130 AID "results tree": an analysis of the number of snowmobiling days in Region III . . . 136 x CHAPTER I STATEMENT OF PROBLEM Introduction Snowmobiling was one of the fastest growing recreation activities in the United States and Canada during the period 1962 to 1970. Sales increased from 4,000 in 1962 to approximately 350,000 in the 1969-1970 season, when this study was conducted. The snowmobile industry expected one million snowmobiles to be in oper­ ation by the 1969-70 season.*- Sales figures for 1969-70 represented a threefold increase compared to the 1967-68 season. This growth appeared even more spectacular when it is realized that 75 to 80 percent of these sales were in the states of Maine, Michigan, Minnesota, New York, 2 and Wisconsin. The state of Michigan had the largest Harold K. Howe, "Industry's View of Problems that Face Snowmobiles," Proceedings of the International Snowmobile Conference (Albany, N . Y . : New York State Conservation Commission, May 1969), p. 48. 2 Steve F. Briggs II, "A Look at Where Industry Is Going," Proceedings of the International Snowmobile Con­ ference (Albany, N.Y.: New York State Conservation Commission, May 1969), p. 51. 1 2 number of snowmobiles at the end of the 1969-70 season, with over 130,000 registered with the state. This rapid growth placed increased pressure on public and private resources suitable for this leisure time pursuit, par­ ticularly on the southern fringe of the "snowbelt" where large urban centres such as Chicago, Detroit, and Toronto have been located.^ This pressure on space and facilities, resulting from the pursuit of the wide variety of activities desired by snowmobilers, was felt at all levels of the public sector: city, county, state, and federal. Conflicts between snowmobilers and other recreationalists, such as skiers, ice fishermen, and cottage owners, have made it increasingly difficult to plan for multiple use of the resources. The problem has been particularly acute near urban centres where frequently the lack of adequate snowcover and comparative scarcity of public open space accentuated the problem. However, despite rapidly growing snowmobile usage, and many associated problems, very little research has been carried out to determine the basic characteristics of snowmobiling and The "snowbelt" refers to that area of northern U.S. which has an average annual snow cover of 100 days or more, i.e., snow of a depth of one inch or more. The northern portion of these five states is in this snowbelt. 3 snowmobilers. Such research would be most useful to planners in allocating resources and in developing man­ agement techniques.^ In Michigan, the problems presented by and con­ fronting snowmobilers have been accentuated by the dis­ tribution of the population and certain environmental factors. Nearly 90 percent of its almost 9 million people live in the southern third of the state while only some 3 1/2 percent live in the Upper Peninsula. 2 As might be expected, the proportion of open space and Only two completed studies existed cvt the time this project was started. They were: (1) "Evenrude Snow­ mobile Trail Survey, 19 67," Outboard Marine Corporation, a mail survey of Evinrude Snowmobile owners to determine the trail designs most desired by the respondents, and (2) "1970 Snow Goer Consumer Survey," Snow Goer Magazine, Minnesota, which mailed questionnaires to a sample of Snow Goer subscribers, primarily to determine the economic impact of snowmobiling. At the time this study was conducted there were three related studies in progress: (1) the, Minnesota Snowmobile Survey - 1970, Minnesota Department of Conser­ vation, which employed a mailed questionnaire to a sample of snowmobile owners to determine the amount and distri­ bution of snowmobile use on public and private lands; (2) 19 70 Snowmobile Survey, Directional Marketing Company, Grand Rapids, which involved a telephone survey of 500 randomly sampled registered snowmobile owners in order to examine travel patterns and economic factors of snow­ mobiling in Michigan; and, (3) An Analysis of Snowmobiling in Ontario, 1969-70, Department of Tourism and information, Province of Ontario, which mailed questionnaires to a sample of registered snowmobile owners to determine the socio-economic characteristics, activity patterns, and economic aspects of snowmobiling in Ontario. 2 U.S. Department of Commerce, Bureau of the Census, 1970 Census of Population, Michigan (Washington, D.C.: Government Printing Office, 1971), p. 3. 4 publicly owned land increases as one moves north and then west through the state. The topography is primarily rolling sandy and gravelly hills with some rocky uplands located in the western part of the Upper Peninsula. The state contains dozens of sizable river systems with a multitude of small streams and lakes.'1' In terms of climate, the most important factors for snowmobilers has been the amount of snowfall and winter temperature patterns. In the highly urbanized areas around Detroit the snowfall was between thirty and forty inches in the 1969-70 season whereas in the upper part of the Lower Peninsula and in the Upper Peninsula 2 many areas received well over 100 inches. This dif­ ference in the supply of snow was further accentuated by the fact that the snowfall frequently melted soon after falling in the southern regions while in the north it tended to accumulate providing a more suitable base for snowmobiling. Because of the abundance of natural resources for recreation, especially in its northern areas, Michigan has become a "recreation state" and tourism is a major "^Department of Natural Resources, Michigan Recreation Plan, 1970 (East Lansing: Michigan Department of Natural Resources, 1970), p. 2. 2 Michigan Department of State Highways, "Snow­ fall Contour Map, Winter of 1969-70" (East Lansing: Department of State Highways, Local Government Division, 1970). 5 part of the economy. To facilitate access to these resources by both Michigan residents and out-of-state recreationalists, an extensive highway system has been developed which has made it easy for most Mid-West residents. The Problem The topic upon which this dissertation focused was that of investigating the use patterns and character­ istics of registered snowmobile owners in Michigan. Since no previous studies of this kind had been done concerning snowmobiling, the question of what were the most appropriate methods to use was an integral aspect of the problem. As Michigan is composed of three rather distinct geographic regions (the Upper Peninsula, the northern Lower Peninsula, and the southern Lower Peninsula), it was of particular interest to determine if there were differences between the snowmobilers in these three regions, and if so, in what ways they differed. This was particularly true in a state such as Michigan since much of the planning of resources and resource management done by the Department of Natural Resources was conducted on a regional basis, using these three geographic divisions. Not only was it of concern to be able to describe snowmobilers in the state with greater accuracy, but it 6 was also important to examine the relationships between their socio-economic attributes and the extent and manner in which they used their snowmobiles; as Van Doren and Lentnek point out: "In order to plan outdoor recre­ ational facilities to meet future requirements of a large urbanized population . . . it is desirable to know the social and economic characteristics of persons engaging in specific activities."'*' If meaningful relationships did exist between user characteristics and use patterns, then it would be useful to determine to what extent use could be predicted from a knowledge of these relation­ ships. This kind of information would be useful to planners who have been faced with making decisions regarding the allocation of natural resources for recreational purposes. Also those persons and agencies concerned with developing policies regulating the use and management of public lands for recreation were in need of this kind of information to provide them with a better understanding of the clientele served. The problem selected for investigation in this dissertation is the question whether or not significant differences in use patterns and socio-economic charac­ teristics existed between snowmobilers from Michigan's "*"Carlton S. Van Doren and Barry Lentnek, "Activity Specialization Among Ohio's Recreation Boaters," Journal of Leisure Research, 1 (Autumn 1969): 296-315. 7 three regions, and if so, of what significance were these differences in managing resources for snowmobiling. Objectives The following objectives were established based on the problem outlined above: 1. To determine and compare socio-economic and snow­ mobile ownership characteristics and attitudes toward snowmobiling regulations of snowmobile owners resident in each of the three regions and the state as a whole. 2. To determine and compare use patterns of snow­ mobile owners for each of the regions and the state as a whole. 3. To establish ratios between total amount of use and amount of use in the county of residence for each region. 4. To examine relationships between the total amount of snowmobile use in each of the three regions and socio-economic characteristics and other selected factors of snowmobile owners resident in each of these three regions. Significance of the Study The tremendous increase in the number of snow­ mobiles in the short space of eight years just prior to 8 the conduct of this study has been cause for some concern / to planners and managers of public recreation land. As Dodge, of the Parks Division, Michigan Department of Natural Resources, pointed out: "Among current problems concerning State Park administrators has been increased pressure from recreation vehicle owners."'*' The fact that two International Snowmobile Con­ ferences (at Albany, New York in May 1969 and at Duluth, Minnesota in February 1970) have been held and that a third (at Portland, Maine in October 1970) was being planned at the time this study was conducted, indicated the desire on the part of planning agencies, management agencies, manufacturers, dealers, and participants to resolve some of the problems that had arisen. Snowmobil­ ing was not a fad that would disappear in the near future but rather as Koenings, of the Bureau of Outdoor Recre­ ation, pointed out at the first of these conferences, " . . . snowmobiling is a legitimate use of peoples' leisure time in the winter, and, therefore, should be 2 provided for." Providing for snowmobiling is not simply a matter of permitting snowmobilers to use public land, for aside "^Robert O. Dodge, December 1970, p. 45. 2 "Michigan," Parks and Recreation, Roman H. Koenings, "Introduction," Proceedings, International Snowmobile Conference (Albany, New York: May 1^69), p. 2. 9 from the impact these vehicles have on the physical environment, there are also problems with conflicts. As Baldwin indicated, such conflicts are: . . . inevitable between off-road vehicle proponents and more traditional outdoorsmen. Left alone, how­ ever, vehicles quickly dominate the scene, since competition of uses is inherently unequal. The , same unequal competition holds true for snowmobiles. As well as resolving conflicts with other recreationists, recreation managers have another task according to Glasgow, Interior, namely: former Assistant Secretary of the " . . . to learn how to so regulate and conduct the use of snowmobiles that the advantages of this type of transportation may be employed and engaged but without degrading the landscape and the natural resources which make up the environment. . . . " 2 Much has been said and written about snowmobiles and snowmobile operators, but little research has been carried out to examine the many aspects of this recent leisure phenomenon. The Michigan Department of Natural Resources has begun to experiment with several schemes for coping with the snowmobile onslaught, by setting up "snowmobile demonstration areas" to test types of trails ^"Malcolm F. Baldwin, The Qff-Road Vehicle and Environmental Quality (Washington, D.C.: The Conservation Foundation, 1970), p. 26. 2 Leslie L. Glasgow, "Snowmobiles Today," Trends in Parks and Recreation, 6 (October 1969): 3. 10 and facilities and through the zoning of areas in the state parks, in an attempt to improve their ability to provide for all winter recreationists who wish to use the state parks.^ These have been important and necessary steps but more knowledge of snowmobilers and their behavior was needed if planners were to plan effectively. In discussing the changing nature of planning, Burton and Cherry emphasized that planners must develop a better understanding of the total situation and in so doing need to rely on the social sciences as well as the physical sciences to develop these insights. 2 In order to under­ stand a situation, or a behavioral phenomenon, it is necessary to examine the relationships which exist. It Was not useful enough to merely measure participation, for as Driver and Tocher pointed out in their discussion on planning in recreation: "We make estimates of short run participation rates and selected recreational activ­ ities, and these are too frequently taken as demand projections." 3 ■^Dodge, "Michigan," p. 45; Michigan Department of Conservation, Snowmobile Demonstration A r e a s , February 15, 1968, p. 2. 2 T. L. Burton and G. E. Cherry, Social Research Techniques for Planners (London: George Allan and Unwin Ltd. 1970), pp. 6-7. JB. L. Driver and S. Ross Tocher, "Toward a Behavioral Interpretation of Recreational Engagements, with Implications for Planning," Elements of Outdoor Recreation Planning, ed. B. L. Driver (Ann Arbor: University Microfilms 1970) , p. 27. XI In order to plan and develop areas for snowmobiling and also to manage the resources used, it was necessary to develop a deeper insight into the nature of this leisure pastime. This study will provide some of the basic data needed for snowmobile facility and resource planning. It will also provide a data base upon which future studies can be founded, resulting in a greater understanding of snowmobilers and also the trends that have been taking place in the pursuit of this activity. Definitions The following terms and definitions were used in this dissertation: Snowmobile.— "Snowmobile means any motorized vehicle designed for travel primarily on snow or ice, steered by wheels, skiis, or runners. O w n e r .— "'Owner* means any person, other than a lienholder, having the property in or title to a snowmobile entitled to the use or possession thereof." 2 Registered O w n e r .— A "registered owner" was defined as the owner of a snowmobile who had registered his vehicle with the Secretary of State. 1 . Michigan, An Act to Register and Regulate Snow­ mobiles , H.B. 3575, Regular Session, 74, Legislature, 1968, Section 1(e). ^Ibid., Section 1(b). 12 D a y .— Respondents to the questionnaire were asked to count each day or part of the day spent snowmobiling as ONE day. This meant that if use of their snowmobile involved several different activities in one day then they would count their time spent on each activity as one "day" for that activity. Head of Household.— "Head of the household" was considered to be the principal wage earner or the person who was considered by the household in question to be head of that household. Members of the Household.— Members of the house­ hold were those persons living in the dwelling unit in question and was usually comprised of members of the "heads" family. Socio-economic Characteristics.— The socio­ economic characteristics selected for this study were as follows: age, educational level, and occupation of household head; combined household income, number of children in household, membership in snowmobile clubs and snowmobile groups preferred. Predictor Variable.— The term "predictor" variable was used in terms of accounting for some portion of the variation in the dependent variable and not in terms of being part of a functional relationship. 13 Limitations and Assumptions The main limitation encountered in conducting the study was one of gaining sufficient financial support to obtain an adequate sample size on a county-by-county basis. The funds available limited sample size to a maximum of 5,000 registered snowmobile owners. There­ fore, the state was stratified into only three geographic regions when designing the survey. The universe from which the sample was selected was limited to those snow­ mobile owners who had registered a machine with the Secretary of State's office by April, 1970. It was assumed that a high percentage of the state's snowmobile owners (other than those who used their snowmobiles exclusively on their own land) would have registered their machines by this date. The Act requiring regis­ tration went into effect on January 1, 1969, providing up to fifteen months during which registration could have taken place. Another limitation was the recall problem result­ ing from asking questions concerning snowmobiling done during the season. Some respondents probably over­ estimated use while others may have underestimated it. Whether or not statistical results were biased in one direction or the other is not known. However, it was assumed that the statistical results were not affected 14 to any great extent; that is, the plus or minus dis­ crepancies in estimating the amount of use added to zero or close to it. Having defined the problem and stated the objec­ tives the next task was to determine the most appropriate methods and procedures to use to investigate the problem, within the limitations outlined. Outline Chapter II is divided into two sections: the first discusses and describes the methods that were used to collect data, and the second describes the methods of analysis. In the first section of Chapter III the results of the contingency tables are presented and analyzed; in the second portion the interaction between the variables is examined. Chapter IV contains the conclusions and recommendations. CHAPTER II METHODS AND PROCEDURES This study was undertaken in cooperation with the Recreation Resource Planning Division,^ Michigan Department of Natural Resources and the Recreation Research and Planning Unit, Department of Park and Recreation Resources, Michigan State University. It was begun in the fall of 1969 but the actual mail survey was conducted in the spring of 1970. In selecting a design for this study, a number of major constraints were taken into consideration. Various possible methods were examined in order to determine their relative suitability under these con­ straints . Data Collection Techniques In this study, limited financial resources was one of the chief constraints. Its main influence was on sampling procedure and the method of collecting data. ■''Now, Office of Planning Services. 16 In examining the various possible approaches, the impor­ tance of keeping survey bias and errors to a minimum had to be kept in mind. Several methods of data collection that warranted consideration were: direct observation, personal inter­ views, telephone interviews, self-administered question­ naires, and examination of existing records. Direct observation as a method did not offer the scope desired. As Burton and Cherry pointed out: "It is usually suitable for only a small fraction of the subjects the researcher wants to study, since it is con­ fined in time and space."^ Snowmobiling activity usually takes place over a wide geographical area and often has undetermined starting and ending points, thus making objective and comprehensive observation very difficult 2 to achieve. In addition, observation would require a considerable field staff and could not provide data on socio-economic characteristics. Collecting information through personal inter­ views had a number of advantages that made this approach worth consideration. Some of the more pertinent ones ■^Burton and Cherry, Social Research Techniques, p. 126. 2 L. L. Lanier, "Snowmobile Survey of Selected State Parks" (unpublished mimeographed material, Recreation Research and Planning Unit, Michigan State University, 1970). 17 which have been discussed by Lininger and Warwick are that this method: (1) tends to reduce the problem of nonresponse common to other survey techniques, (2) is more readily administered to persons of all education levels, (3) allows the interviewer to correct misinter­ pretation of questions, (4) facilitates exploration of areas in which little information existed, and (5) makes it easier to obtain information about emotionally charged subjects.^ However, a number of problems would arise in using the personal interview approach on a state-wide basis, the most important of which would be the high cost of conducting interviews throughout the state, especially if numerous "call-backs" were required in order to make contact with the potential respondents. The administrative structure to carry out interviews in different portions of the state would have added considerably to the cost. According to Seltiz et al., telephone interviews offered some advantages compared to the mailed question­ naire, primarily, that they were usually less costly per unit. However, this technique also had serious limitations for this study, such as: (1) having to be brief and superficial in order to gain the cooperation ^Charles A. Lininger and Donald P. Warwick, "Intro­ duction to Survey Research" (Ann Arbor: Survey Research Center, Institute for Social Research, University of Michigan, mimeographed material, 1967), pp. IV - 2 and 3. 18 of the respondent, (2) not being able to reach a random sample of the desired population, since not all people had telephones, and (3) the difficulty in contacting persons who work away from home by telephone.^ Burton and Cherry noted that with the mailed self-administered questionnaire: 11 . . . it is possible to cover a wider geographical area and to reach a larger sample of the population (with given financial resources) 2 than is possible by the use of an interview survey." According to Lininger and Warwick, the self-administered questionnaire had several other advantages compared to the interview method, such as: a feeling on the part of the correspondent that his statements would be treated confidentially making it easier to express his true views, and that he had greater opportunity to search out information which he may not have had at his fingertips. 3 Against these advantages there were some limitations that needed evaluating. As Moser pointed out: "The vital limitation of mail survey is the difficulty of ^Claire Selltiz, et al., Research Methods in Social Relations (New York: Holt"^ Rinehart and Winston, 195*!)) , p. 239. 2 Burton and Cherry, Social Research Techniques, p. 38. 3 . Lininger and Warwick, Survey Research, pp. IV 1 and 2. 19 getting an adequate response."'*' He also indicated that questions must be kept simple in order to be understood with the help of printed instructions, that the answers obtained must usually be accepted as final, and that no opportunity was available to supplement the respondent's answers by observational data. The advantage of using existing records as a source of information is that the data have already been collected. However, since such records have usually been produced for some other purpose they may not be in an appropriate form. After carefully considering the above methods, while keeping the objectives of the study in mind, it was decided to use a combination of several methods. The self-administered questionnaire was thought to be the most appropriate method for collecting the main body of infor­ mation about snowmobilers. in selected counties, In order to determine if, there was a difference in the characteristics of the respondents compared to non­ respondents, a telephone survey of a sub-sample of both nonrespondents and respondents in those counties was conducted. Registration records from the Secretary of State Office were used to select the survey sample, and snowfall data collected by the Michigan Department of ■*"C. A. Moser, Survey Methods in Social Investigation (London: Heinemann Education Books Ltd., 1958) , pp. 177-78. 20 State Highways was used to estimate the snowfall received in each county. The rationale underlining these selections will be discussed more fully in the following sections. Self-Administered Questionnaire The chief reason for selecting this method of data collection was because it enabled the researcher to con­ tact a larger sample than would have been possible with the same amount of funds if the personal interview method had been chosen. Due to the larger sample it would be possible to make more precise statements about snowmobilers throughout the state. Also, data of a more comprehensive nature could be gathered than if a tele­ phone survey was used. Having made that decision, the problem of developing a suitable survey instrument had to be tackled. of decisions, Oppenheim identified three other types that are pertinent to the study, which had to be made before writing the actual questions, namely: (1) The method of approach to the respondents (after selection through sampling procedures), including sponsorship, stated purpose of the research, confidentiality, anonymity. (2) The build-up of question sequences in the order of questions and other techniques was in the frame­ work of the questionnaire. (3) The use of precoded versus free-response questions.1 A. N. Oppenheim, Questionnaire Design and Atti tude Measurements (New Yorkl Basic Books, Inc., 1966), pp. 24-25. 21 Personnel from the Recreation Resource Planning Division were of the opinion that it would be better to mention Michigan State University as the sponsoring agency since it was said to have a better image of impartiality than the Department of Natural Resources at that time.''' According to Burton and Cherry: "The initial response to a self-administered survey depends upon . . . the degree of interest in the subject of the survey that 2 can be aroused in this population. . . . " Hence, a covering letter for inclusion with the questionnaire was drawn up which asked for the respondent's cooperation. It said that the survey was being conducted in cooper­ ation with the Department of Natural Resources, briefly explained the purpose of the study, stated that the respondents' replies would be treated confidentially, and indicated that it would only take fifteen to twenty minutes of their time (see Appendix A ) . By assuring confidentiality rather than guaranteeing anonymity, it was possible to place each respondent's registration number on the questionnaire so that it could be checked off on a duplicate set of address labels if the "'‘A series Colburn and other Planning Division that the covering 2 p. 39. of discussions were held with Mr. William members of the Recreation Resources in early March 1970 regarding the forms letter and questionnaire should take. Burton and Cherry, Social Research Techniques, 22 questionnaire was returned. This procedure reduced the number of reminder cards and second questionnaires which had to be mailed out subsequently. In developing the overall plan for a question­ naire, Oppenheim suggested that the first part should begin with: " . . . some easy impersonal questions and not ask for details like age, family, occupation and so forth until rapport has been established."^ In deciding on the type of questions to use (i.e., open-ended versus closed question), it was decided to use closed questions where the respondent was either asked to check off the appropriate category or he was asked to fill in a blank with a name or a quantity. Cherry stated: As Burton and "Open questions are easy to ask, dif- fxcult to answer and more difficult to analyze." 2 In some cases the category "other" was included and the respondent asked to specify the particular activity, etc. This was done to ensure that, at least in a sense, the categories listed were exhaustive in nature. Special efforts were made to ensure that the wording of the question was simple and straight forward. In order to reduce the length of the questionnaire, charts were constructed that enabled the respondent to put down ^"Oppenheim, Questionnaire Design, p. 37. 2 Burton and Cherry, Social Research Techniques, p. 57. 23 a considerable amount of information in a small space. This technique gave the questionnaire the appearance of being shorter than if several questions had been asked to obtain the same amount of information, for as Crapo and Chubb concluded: "The amount of material contained in a questionnaire does not seem to affect response as much as the appearance of being b r i e f ."^ The questionnaire was pretested in early April on thirty-six snowmobile owners who either worked in a nearby manufacturing plant or were civil servants. As a result of the pretest, minor but significant changes were made in the wording of the questions and also in the categorization in some cases. The average time required to complete the questionnaire was less than twenty minutes. The revised questionnaire requested the following information: (1) Information describing the snowmobile make, year of manufacture, horse-power, number of years owned, county in which it was registered, and the member of the household who was the registered owner (see question 1, Figure 1). Douglas Crapo and Michael Chubb, Recreation Area Day-Use Investigation Techniques: Part 1 A Study of Survey Methodology, Technical Report Number 6 (East Lansing, M i c h .: Recreation Research and Planning Unit, Michigan State University, 1969), p. 97. SN O W M O B ILE M IC H IG A N USE STUDY 2 1 DO YOU FEEL THAT REGULATIONS SHOULD BE ESTABLISHED TO CONTROL SNOWMOBILE ACTIVITY ON FROZEN LAKES WHERE ICE-FISHING IS BEING DONE? 22 i— I YesLJ . ri Noli DO YOU THINK SNOWMOBILERS SHOULD BE ALLOWED TO DRIVE ON A: (a) main highway (other than expressways)? YesQ (b) secondary highuavs? Yes (3 (c) highway shoulders (unplowed portions)? Yes □ n° Q (d) street of a town or village? YesQ No Q 1 DESCRIBE THE SNOWMOBILES OWNED BY MEMBERS OF YOUR HOUSEHOLD, BY COMPLETING THE TABLE BELOW. (Head of household means the nain wage earner.) No ( 3 Make EXAMPLE 23 Horse­ power Year No. of years owned County of registration a 11*7 Registered owner, eg., head, wife, son, etc.. d / ft * * * - IN THE SPACE BELOW, PLEASE INDICATE ANY SPECIAL SNOWMOBILINC PROBLEMS YOU MAY HAVE HAD DURING THE PAST SEASON. PLEASE CHECK THE FOLLOWING ITEMS THAT ARE OWNED BY MEMBERS OF YOUR HOUSEHOLD. 3 Motorcycle or trail bike □ Snowmobile conversion kit (for summer use) □ All terrain vehicle □ Snowmobile trailer (to carry snowmobile) Q Truck camper □ Power boat F~| Camping or house trailer (that you tow with your car) Q WHICH COUNTIES DID YOU USE THE HOST FOR SNOWMOBILING DURING THE PAST WINTER? (Write in the number of days on each line. NOTE: Count each day or part day spent snowmobiling in a county as ONE day.) County of most use County of 2nd most use County of 3rd most use All other county use Out of State use; name state or Prov. County name No. of days of use EXAMPLE X County name No. of days of use THANKS FOR YOUR HELP! £ 7 /& JL t IN WHAT YEAR DID YOU BUY YOUR FIRST SNOWMOBILE? If you accidently misplaced the return envelope provided, please mall to: Recreation Research and Planning Unit Room 131 Natural Resources Building Michigan State University East Lansing, Michigan 48823 1 (a) IS THERE A SNOWMOBILE CLUB IN YOUR AREA? Yes [ 3 (b) DO YOU BELONG TO A SNOWMOBILE CLUB? Yes Q No □ No □ Don't know I I Fig. 1. The snowmobile questionnaire. A view of the front and back pages of the questionnaire in the unfolded condition. Actual size of each page was 8 1/2 x 11 in. 6 INDICATE BELOW THE KINDS OF AREAS THAT YOU USED FOR SNOWMOBILING DURING THE PAST SNOWMOBILE SEASON, BY WRITING IN THE NUMBER OF DAYS FOR EACH OF THE AREAS BELOW. (NOTE: Count each day or part day spent in an area as ONE day. Therefore the total days may add to more than the actual number of days you used your snowmobile.) On your own land _____ days On private land after paying a fee On Federal land _____ days On local public roads not plowed On State owned land On county owned land days days On private land, at no charge days 14 ON HOW MANY DAYS DID YOU OR MEMBERS OF YOUR HOUSEHOLD TAKE PART IN THE FOLLOWING ACTIVITIES WHILE USING YOUR SNOWMOBILE(S) DURING THE PAST SNOWMOBILE SEASON? (Write in the number of days on each line. Use a "0" for no activity. NOTE: Count each day or part day spent on an activity as ONE day.) Scrambling in open areas and on lakes _____ days On City park land (including City golf courses) days Trail riding and forest Snowmobiling to work or during your vork • days Other cruising _____ days _______________________ (specify) days davs (specifv) 15 ON HOW MANY DAYS DID YOU OR MEM3ERS OF YOUR HOUSEHOLD ENCAGE IN THE FOLLOWING ACTIVITIES WHILE ON A SNOWMOBILE TRIP THE PAST SNOWMOBILE SEASON? (Write in the number of days.) Tobogganing, sledding or skiing (not being towed) v ft Hunting 6 Ice fishing 7 Competive racing days ----- Cook-outs , , Overnight camping „ , Other __________________________ (specify) davs • davs days days , days days WHERE IS YOUR PERMANENT RESIDENCE? county state 8 WHAT IS THE AGE AND SEX OF THE HEAD OF YOUR HOUSEHOLD? 9 WHAT IS THE OCCUPATION OF THE HEAD OF YOUR HOUSEHOLD? Age ____ years zip code MaleQ FenaleQ 1 6 HOW MANY NIGHTS DID YOU STAY OVERNIGHT AWAY FROM HOME WHILE ON A SNOWMOBILE TRIP DURING THE PAST SNOWMOBILE SEASON? (Complete the table below by writing the number of trips opposite the distance travelled.) Distance from heme to area occupation (not organization) Number of TWO night trips Number of ONE night trips Number of THREE or more night trips Up to 50 miles 10 WHAT IS THE AGE AND SEX OF EACH FAMILY MEMBER LIVING IN YOUR HOUSEHOLD?(Not head of household) Male - ages ___ : ___ ; ; ___ ; . Female - ages ___ ; ___ ; ___ ; ; 100 to 200 miles 200 to 300 miles 1 1 WHICH OF THE ANSWERS BELOW BEST INDICATES THE TOTAL YEARS Or EDUCATION COMPLETED BY THE HEAD OF THE HOUSEHOLD? (Check one) □ □ □ □ □ □ □ □ □ □ 4 5 6 7 8 9 10 11 12 Over 300 miles □ □ □ □ 13 14 15 16 17 or more 17 under $3,000 □ S8.000 - 59,999 □ 520,000 - 524,999 $3,000 - $5,999 □ $10,000 - 514,999 □ S25,000 - S29.999 □ $6,000 - 57,999 Q $15,000 - $19,999 □ $30,000 and over □ DO YOU FEEL THAT THE PRESENT REGULATIONS SHOULD BE: □ much more strict more strict □ WHICH OF THE FOLLOWING BEST DESCRIBES THE COMBINED GROSS INCOME OF YOUR HOUSEHOLD IN 1969? (Check one) □ 12 to Ln 50 to 100 miles unchanged □ less strict □ much less strict □ 18 DO YOU FEEL THAT THE ENFORCEMENT OF THE PRESENT REGULATIONS SHOULD BE: □ much more strict □ more strict □ unchanged □ □ less strict much less strict 13 CHECK THE FOLLOWING PEOPLE OR GROUPS OF PEOPLE THAT YOU WENT SNOWMOBILING WITH MOST OF THE TIME DURING THE PAST SNOWMOBILE SEASON. I went alone □ A Wife (or husband) □ An My children d (Check only one) group of friends Q organized group 0 19 SHOULD CHILDREN UNDER 14 YEARS OF AGE BE ALLOWED TO OPERATE A SNOWMOBILE? Family (with children) Q (a) by themselves without adult supervision? Yes LI *>□ Girlfriend (or boyfriend) Q (b) only under the supervision (within sight) of an adult? VesQ No|_j Other ___________________ (specify) □ 2 0 D0 YOU THINK THAT REGULATIONS SHOULD BE ESTABLISHED TO GOVERN THE AMOUNT OF NOISE THAT A SNOWMOBILE IS ALLOWED TO MAKE? f— i .pi Fig. 1. Continued. The inside pages 26 (2) Ownership of selected types of recreation equip­ ment, particularly items that were motorized or related to motorized vehicles (see question 2, Figure 1). (3) The amount of snowmobiling done in the 1969-70 season, measured in "days." The respondents were asked to specify the three most frequently used counties plus the number of days of all other county use (see question 3, Figure 1). (4) The number of years the registered owner had owned a snowmobile and whether or not they belonged to an organized snowmobile club (see questions 4 and 5, Figure 1). (5) The type of land used, classified according to ownership, and the number of days used, in each case, for snowmobiling (see question 6, Figure 1). (6) The socio-economic characteristics of the house­ hold (see questions 7-12, Figure 1). (7) The kinds of groups with whom snowmobile owners most preferred to go snowmobiling (see question 13, Figure 1). (8) The kinds of activities pursued by the owner or members of the household while using the snowmobile(s) and also those activities 27 associated with a snowmobile trip. In each case the number of days involved was requested (see questions 14 and 15, Figure 1). (9) The number of one-, two-, and three-night snow­ mobile-oriented trips taken during the season and the distance travelled from the place of residence (see question 16, Figure 1). (10) The attitude of owners towards selected regu­ lations related to the operation of a snowmobile (see questions 17-22, Figure 1). (11) The last question was the one open-ended question in the questionnaire and it provided space for the snowmobiler to mention any special snowmobile problems he may have had during the past season (see question 23, Figure 1). Sample Design Moser stated that: lie all sample design. "Two major principles under­ The first is a desire to avoid bias in the selection procedure, the second broadly to achieve the maximum percision for a given outlay of resources."^ In order to avoid selection bias it was necessary to use a probability sampling method of choosing the sample and to make sure the sampling frame adequately covered the ^Moser, Survey Methods, p. 73. 28 population to be studied. Moser went on to say that: "Stratification is a means of using knowledge of the popu­ lation to increase the representativeness and the pre­ cision of the sample."^ In Michigan there is a transition from very densely populated areas in the southern portion of the state section) (and particularly in the southeastern to very sparsely populated areas in the north and northwest portions. In this continuum from a very urban population to a very rural one it was known that there were also changes in the characteristics of the population in terms of income, education, occupation, and other socio-economic characteristics. Whether dif­ ferences in the characteristics of snowmobile owners occurred from one area to another in a similar manner was of interest in this study. It was also apparent that variation in snowfall and open space availability could result in quite different use patterns among snowmobilers residing in different areas. Therefore, it appeared that some additional precision could probably be gained by stratifying the sample on a geographical basis. The problem now became one of determining appro­ priate strata. Data on a county basis rather than on a broader regional basis would be more useful to planners. This was particularly true in the case of Michigan since it varies considerably in terms of population distribution, ^"Ibid., p. 78. 29 topography, climate, snowfall, accessibility, etc. Therefore, reliable data on a county-by-county basis would more clearly indicate how the characteristics in use patterns of snowmobilers varied. In other words it would give planners more detailed and complete infor­ mation upon which to make policy decisions. To obtain data of this type, a sample of approxi­ mately 24,000 snowmobilers would be required in order to insure a return of some 200 responses per county for each of Michigan's 83 counties.^ However, available funds only allowed for a sample of approximately 5,000. It was, therefore, decided to stratify the state into three regions corresponding to the regional division established by the Parks Division, Department of Natural Resources: Region I, the Upper Peninsula; Region II, the Upper Lower Peninsula; Region III, the Lower Lower Peninsula (see Figure 2). According to Burton and Cherry in their discussion of sample size: "A variable sampling fraction can greatly increase the accuracy when the sampling units vary greatly in size, or more generally in the variability In discussions with Dennis Gilliland, Associate Professor, Department of Statistics and Probability, it was estimated that in most cases a minimum of 200 returns would be necessary in order to make inferences about the snowmobile population of a county. A sample size of 24,000 assumed a response rate of 65 percent. 30 IIU ftOTALt arquet Sc h o o l c r a f t t tSC*A*9» REGION I ,oru«o'hobtkwi., M i J » . i a | c , ^ „ OIK1, o « c o D . *°»co».' o tii A* I IO IC O .rNh-i...]. Tl»«I r®j««fi*rCL*tl *«l.»OWI»T REGION II IC I« N > , « » » * • < > ^M C C O S T A I e/rr flCOL ONTCALM ¥0 t R lT lO T I 1 *LAPit iC U N T O N | i M I A * A 99909 9Anp9 REGION III an^ c o « D (v: L 4 lw /* f L L IB A N AN BURCN l BAKOV « \.A N A X .' IA T O N C ALHO UN . JAC R80N 09990!t JAOKtOU I I CAM ■I r . j o l K R N l1 B R A N C H " ! " , L L l M l -e ! t . I N A V r « t 1 ' ] 06834 Fig. 2. Regions and selected counties used in the sample stratification and analysis. from stratum to stratum."'*' Kish stated that: the variance decreases as n increases. . . . " "Generally, 2 Since the size of the snowmobile population varied considerably between each of the three regions, samples from each region based on the same percentage of registered snow­ mobiles in each case would result in a very small sample being selected from the Upper Peninsula and a rather large sample from the southern Lower Peninsula. There­ fore, it was decided to use a disproportionate sampling method which would result in a larger than average sampling fraction from the less densely populated areas. 3 This approach closely approximated what Kish refers to as "optimum allocation" criteria as opposed to proportionate sampling. 4 Since little was known about the characteristics of snowmobilers in Michigan, it was necessary to arbitrarily set the sample size for each region, as: 1,000 from Region I, 1,500 from Region II, and 2,500 from Region III. ^"Burton and Cherry, Social Research Techniques, pp. 104-05. 2 Leslie Kish, Survey Sampling Wiley and Sons I n c . , 1965), p. 92. 3 Gilliland. 4 (New York: John Further discussion. Kish, Survey Sampling, pp. 92-93. Kish uses "optimum allocation11 to refer to the process of obtaining the most meaningful sample sizes under given set of cir­ cumstances rather than the "best" sample size, per se. 32 To insure that some reliable data were obtained on a county basis in each region (in order to permit analysis at this level, if desired), eight counties were selected from which a larger proportion of registrations would be selected: Marquette County from Region I; Bay and Grand Traverse counties from Region II; and Genesee, Ingham, Kent, Oakland, and Wayne counties from Region III. The criteria used in the selection of these counties were snowmobile population, geographic location, and available socio-economic data on a comparable basis (primarily from boating studies conducted by the Recreation Research and Planning U n i t ) . In each case, the sample size was set at 300 registrations per county, assuming that approxi­ mately 200 questionnaires would be returned in each case. The remainder of each region was then to be sampled on a set proportion within that region. The sampling frame was a listing of registered snowmobiles on magnetic computer tape which was obtained from Michigan Secretary of State Office in April, 1970. The Michigan State University CDC 6500 computer was pro­ grammed to select a specified size random sample from each county.'*' Based on the number of registered snow­ mobiles in each county and the above criteria, the ■*"In the selection of each observation, the com­ puter was programmed to call on its "random number generator." This meant that each observation in the sample frame had an equal chance of being selected. 33 computer was instructed to calculate the fraction required to obtain the desired sample size for each of the eight selected counties and the remainder from each of the three regions. The number of registered snow­ mobiles, the number of registrations drawn by the sampling process, and the percentage of the registrations drawn for each county were shown in Table 1. Distribution and Returns As mentioned above, a common limitation of self­ administered questionnaires has been the likelihood of a low response rate. Therefore, every attempt needed to be made to ensure a high rate of return. This study had an advantage in dealing with a population that had a common interest in snowmobiling, a pastime that was relatively new and also had created some concern particularly among nonsnowmobilers, all of which was speculated by the researcher to motivate the recipients of questionnaires to complete and return them. Also included with the questionnaire was an explanatory letter, a map of the state showing the counties and major roads dix A ) , and a prepaid return envelope. (see Appen­ These items were all expected to aid response. In addition to the above measures intended to increase the initial response, methods aimed at securing TABLE 1 SAMPLING STATISTICS County Code No. No. Of Reg. SMa % of Total Requested Sample Drawn Usable Responses % of Total Region I - Upper Peninsula Marquette Alger Baraga Chippewa Delta Dickinson Gogebic Houghton Iron Keweenaw Luce Mackinac Menominee Ontonagon Schoolcraft 52 02 07 17 21 22 27 31 36 42 48 49 55 66 75 Total 3,762 970 565 2,725 2,238 1,321 1,052 1,343 1,012 91 778 979 1,049 1,075 805 8.0 4.4 11 II 11 II II It II IV 11 II II II If 19,765 319 47 25 118 101 48 45 57 41 3 34 48 34 41 38 221 30 16 63 70 39 30 35 29 2 22 32 22 25 29 5.9 3.1 2.8 2.3 3.1 3.0 2.9 2.6 2.9 2.2 2.8 3.3 2.1 2.3 3.6 999 665 3.4 184 168 14 43 31 4.6 6.2 2.8 2.5 3.0 Region II - Upper Lower Peninsula Bay Grand Trav. Alcona Alpena Antrim 09 28 01 04 05 3,978 2,700 505 1,741 1,038 7.5 11.2 3.2 VI II 276 292 18 62 42 TABLE 1— Continued County Arenac Benzie Charlevoix Cheboygan Clare Crawford Emmet Gladwin Iosco Isabella Kalkaska Lake Leelanau Manistee Mason Mecosta Midland Missaukee Montmorency Newaygo Oceana Ogemaw Osceola Oscoda Otsego Presque Isle Roscommon Wexford Total Code No. 06 10 15 16 18 20 24 26 35 37 40 43 45 51 53 54 56 57 60 62 64 65 67 68 69 71 72 83 No. of Reg. SMa 976 545 1,307 1,509 1,126 424 1,073 723 1,168 1,366 578 222 622 687 818 1,042 1,487 546 481 1,264 1,112 1,065 816 223 1,137 915 1,314 1,198 35,706 % of Total Requested II II II II II II II II II II II II II II II II II 3.2 II II II II II II II II II II Sample Drawn Usable Responses % of Total 40 12 33 51 31 13 37 24 51 52 20 2 14 18 24 32 53 16 7 30 39 38 27 6 45 30 46 41 31 7 25 47 23 11 26 19 36 40 20 1 23 15 13 22 38 12 5 20 27 17 21 7 28 21 36 29 3.2 1.3 1.9 3.1 2.1 2.6 2.4 2.6 3.1 2.9 3.5 0.5 3.7 2.2 1.6 2.1 2.6 2.2 1.0 1.6 2.4 1.6 2.6 3.1 2.5 2.3 2.7 2.4 1,522 1,060 3.0 TABLE 1— Continued County Code NO. No. of R eg. SMa % of Total Requested Sample Drawn Usable Responses % of TotalJ Region III - Lower Lower Peninsula Genesee Ingham Kent Oakland Wayne Allegan Barry Berrien Branch Calhoun Cass Clinton Eaton Gratiot Hillsdale Huron Ionia Jackson Kalamazoo Lapeer Lenawee Livingston Macomb Monroe Montcalm Muskegon Ottawa 25 33 41 63 82 03 08 11 12 13 14 19 23 29 30 32 34 38 39 44 46 47 50 58 59 61 70 8,098 3,448 4,704 7,107 4,810 1,379 911 1,051 470 1,122 398 1,124 1,332 1,171 346 1,056 1,074 1,714 1,790 1,879 452 749 3,820 593 1,970 2,868 1,623 3.8 8.8 6.4 4.3 6.3 2.3 2.3 II II II II II II II II II II If II II II II II II II It II 340 294 310 298 309 37 23 39 12 24 9 31 24 34 11 30 14 36 48 41 13 20 86 8 53 66 37 224 172 207 219 189 26 22 27 6 15 3 39 27 21 10 21 10 26 28 31 9 15 63 5 37 44 34 2.8 5.0 4.4 3.1 3.9 1.9 2.4 2.6 1.3 1.3 0.8 3.5 2.0 1.8 2.9 2.0 0.9 1.5 1.6 1.7 2.0 2.0 1.7 0.9 1.9 1.5 2.1 TABLE 1— Continued County Code NO. Saginaw Sanilac Shiawasee St. Clair St. Joseph Tuscola Van Buren Washtenaw 73 74 76 77 78 79 80 81 No. of R e g . SMa 5,751 1,848 1,476 1,761 299 2,133 778 1,193 % of Total Requested II II II 2.3 If II II If Sample Drawn Usable Responses % Of Total" 136 43 31 46 10 57 18 24 92 31 27 33 9 50 14 16 1.6 1.9 1.8 1.9 3.0 2.4 1.8 1.4 Total 72,298 2,612 1,802 2.5 TOTAL 127,769 5,133 3,527 2.8 aTotal number of registrations on the magnetic tape obtained from the Secretary of State's Office. ^Usable responses as a percentage of the total number of registrations in that county. 38 additional responses were considered. As Burton noted: "The most well known of these is the follow-up letter, or reminder."'*' to send: Moser pointed out that another approach was "Follow-up requests, enclosing a copy of the questionnaire and covering letter. . . . " 2 The factors of cost and time had also needed to be considered, in terms of how many reminders or follow-up requests should be sent out before the point of diminishing returns was reached or the survey unduly delayed. The snowmobile questionnaires, plus enclosures, were mailed to the selected samples of registrants on May 26th and June 1st (see Table 2). The number of questionnaires returned per day rapidly built to 47 5 by June 4th and then began to decline. As the question­ naires were returned, they were checked off on a duplicate set of labels. On June 12th, the number of returns had decreased to 49 per day and the reminder card (see Appendix B) should have been sent out at this time; how­ ever, since only 3,500 cards had been printed, their mailing had to be delayed until June 16th. increased substantially. Returns On June 26th, when returns ■^Burton and Cherry, Social Research Techniques, p. 40. Results of fish and game surveys conducted by the Michigan Department of Natural Resources indicated that reminder cards produced substantial increases in the number of returned questionnaires. 2 Moser, Survey Methods, p. 182. 39 TABLE 2 DISTRIBUTION OF AND RESPONSE TO MAILED QUESTIONNAIRE Date May 28 June 1 2 3 4 5 8 9 10 11 12 15 16 17 18 19 22 Note: Day Mon Number Mailed 3335 1798 Mon Mon 3500 (cards) Mon Returns 26 241 475 184 365 102 85 64 49 78 37 46 58 206 308 Date June 23 24 25 26 29 30 July 1 2 6 7 8 9 10 13 14 Net sample = 5133 - 129 = 5004 Net returns = 3705 - 178 = 3527 Response rate = 70% Day Mon Mon Mon Number Mailed 2616 (2nd Q) Returns 101 92 83 67 230 227 103 114 170 33 66 36 27 0 32 3705 40 per day had dropped to 67, a second questionnaire with a revised explanatory letter (see Appendix C) was mailed to those who had not yet returned their questionnaires, resulting in a further increase in returns. July 14th was set as a cut-off date by which time 3,705 question­ naires had been received. From the original sample of 5,133 registrations, 129 were deleted due to: unknown addresses, snowmobiles not being used, snowmobiles no longer being owned, because the snowmobiles were demonstrators owned by retailers, or because the owners resided outside Michigan. resulted in a net sample of 5,004 registrations. This Of the total 3,705 returned questionnaires, 178 were rejected due to respondents' refusal to fill out the question­ naire, duplicate returns, and residence in another state. This resulted in a net response of 3,527, or a response rate of 70 percent. Data Coding According to Selltiz: " . . . categorization of complex data is usually done by coders after the data have been collected."^ He went on to discuss the impor­ tance of training the coders to improve the reliability ^Selltiz, Research M ethods, p. 402. 41 of the coding and also checking the accuracy and consis­ tency of the coding throughout the procedure.^ The questionnaires were coded by staff members of the Urban Survey Research Unit, Michigan State Uni­ versity. Data from the questionnaires were transferred to optical scan sheets which were over-printed to facili­ tate the coding process in this study. Every fifth questionnaire was checked-coded, a process in which the data were transferred to another optical scan sheet by a check-coder and the results compared. If errors were discovered, they were corrected and the error was brought to the attention of the initial coder. In cases where the error rate of a coder was found to be excessive (over 4%) all of his questionnaires were recoded. The final error rate for the complete coding operation was estimated to be less than 1.5 percent. transferred to computer cards. 2 Data were then Finally a computer pro­ gram, "Try-Hard," was used to transfer data to magnetic computer tape. In the transfer process the program 1Ibid., p. 405. 2 The error rate was calculated by multiplying the total number of detected errors by five (since only every fifth questionnaire was checked) and dividing this product by five times the number of questionnaires times the total number of columns (229). 42 rejected any responses that were outside a set range of acceptability for each question (in other words elimi­ nating gross errors).^ Nonresponse The importance of obtaining a high response rate in order that more precise inferences concerning the snow­ mobile population can be made has been pointed out in preceding sections. However, since 30 percent did not respond there was still the question of whether or not nonrespondents differed from respondents. As Cochran points out, the most important consequence of nonresponse is that estimates can become biased. 2 There may have been any number of reasons why people did not complete and return questionnaires, such as: having moved to a new address, being away on vacation at survey time, too busy to be bothered, refusal to cooperate for one reason or another, etc. Cherry suggest: respondents, ... " . . . Burton and that interviewers be sent to non­ to obtain some data about their Paul Emmery, "Try-Hard" (East Lansing: Urban Survey Research Unit, Michigan State University, Mimeo­ graphed material, 19 70) . This computer program was pri­ marily designed to clean large amounts of punched card data for analysis. 2 York: William G. Cochran, Sampling Techniques John Wiley and Sons, Inc., 1963), p. 389. (New 43 characteristics so that these may be compared with the characteristics of the respondents."'*' It was decided to use a telephone interview to gather a limited amount of data concerning nonrespondents at a reasonable cost. In pursuing this approach, an abbreviated questionnaire was developed which requested the following data: details concerning the snowmobile, the amount of use by county, what kind of land they used for snowmobiling and how much, socio-economic data, and kinds of activity pursued while snowmobiling dix D ) . (see Appen­ Random samples were drawn of both respondents and nonrespondents from the counties of Ingham and Kent. Two interviewers were trained to carry out this survey. This procedure enabled interviewers to telephone sample respondents in the early part of July and concluded with sample nonrespondents following the cut-off date on July 14. Snow Depth The amount of snow an area received probably had considerable influence on the amount of snowmobiling done in that area, particularly in those counties of the southeastern portion of the state, which received little snow. The Parks Division was under pressure, especially from residents of Region III, to open public land to 1 p. 41. Burton and Cherry, Social Research Techniques, 44 snowmobiling. In response to this pressure and in an effort to minimize environmental damage resulting from snowmobile traffic, the Division set four inches as the minimum snow depth required before snowmobiling would be permitted in selected state parks and recreation areas.^ Therefore, if the number of days that a given depth of snow was available in each area was known, then a meaningful index of the number of possible snowmobiling days could be established. The U.S. Department of Commerce records the number of days that there is a one-inch or more and three inches or more snow on the ground throughout the state. 2 However, data for 1969-70 had numerous gaps, that is, depth of snow was not always recorded every month and further, it was not collected for every county. The Michigan Department of State Highways gathered data on the average amount of snowfall for each county in order to establish the amount of financial aid 3 each county should receive for snow removal. This ‘'"Paul Rearick, Parks Division, Michigan Department of Natural Resources, Discussions held in November 1969. 2 U.S. Department of Commerce, Environmental Data Service, "Climatological Data - Michigan, October 1969 July 1970" (Washington, D.C.: Department of Commerce, 1970) . 3 . Michigan Department of State Highways, Local Government Division, "Snowfall data for 1970" (Lansing, Mich.: Department of State Highways, 1970). 45 information appeared to be consistent in its recording methods and calculations. Therefore, it was decided to use the Department of State Highways data to develop an index of relative availability of snow for snowmobiling in each county (see Appendix E ) . Methods of Analysis Phase I Contingency tables were developed using G9 MSU ACT Program designed for the CDC 3600 computer. These tables are comprised of the frequencies and percentages for each variable for each of the three regions and the state. Wherever it was appropriate means and standard deviations were also calculated. to compare: These tables were used (1) socio-economic characteristics, snowmobile ownership characteristics, (2) (3) attitudes toward regulations governing snowmobiling, and (4) patterns of use of snowmobile owners resident in each region. A one-way analysis of variance was carried out on the quantitative variables that were suitable for this type of analysis (see Appendix F ) . If the overall F test was found to be significant at the .05 level, then Scheffe post-hoc comparisons were conducted to determine which pair or pairs of means the regions) (for each of contributed to this significance. As 46 Hays pointed out, this method was applicable to groups of unequal size and suitable for any comparison.^ An overall chi-square test was carried out on the qualitative nominal and ordinal variables. Maxwell, in discussing contingency tables having more than one degree of freedom, stated, " . . . though a significant 2 overall x would tell us that these proportions were heterogenous a more detailed analysis would be required to decide just where the significant differences lay." 2 The contingency tables that were of interest, here, had from two to eight degrees of freedom, therefore, the William L. Hays, Statistics (Toronto: Holt, Rinehart and Winston, Inc.^ 1963), pp. 483-85. In discussions with Howard Teitelbaum, Research Methodology and Evaluation Specialist, Office of Medical Education, Michigan State University, July, 197 2, the following formula was derived from Hays: /v / ib ± / T (J-l) Fa (MSw) £ n. J /N in which ip was the difference between the sample means to be compared; J was the number of groups (3) in the study; Fa was the value required for significance at the a level (.05) with J-l and N-J degrees of freedom; MSw was the mean square within groups; Cj were the weights assigned to each group in the comparison, such that the sum of the weights equalled zero, in this case weights of 1 and (-1) were assigned; nj was the number in each group being compared. In order for the comparison to be significant the interval could not include zero. 2 A. E. Maxwell, Analyzing Qualitative Data (London: Methuen and Co” L t d ., l£61) , p"I 52. 47 overall chi-square was subdivided into additive com­ ponents. In other words, the degrees of freedom were partitioned.^- If the probability level was greater than .05, it was not considered to be significant. In order to examine the patterns of use more closely, the "amount of use" data (as measured by the number of snowmobiling days in the county of first, second, and third most use) were expanded to produce an estimate of the total number of snowmobiling days for all of the registered snowmobile owners in each county. The expansion was carried out by using weights which were based on the response rates from the eleven subregions (see Appendix G ) . 2 This process provided estimates of the amount of movement into and out of each county by snowmobilers in pursuing their sport. Phase II Analysis problems.— Data from cross-section sur­ veys, because they were usually of a complex nature, frequently presented problems of analysis. Morgan and Sonquist identify the major problems as being: (1) the "^The NONP11 Program was used: "Lancaster's Par­ titioning of Chi-square," designed for the IBM 360/67 computer at the University of Alberta, Canada, 1971. 2 The weights used were the reciprocal of the proportion that the number of usable responses was of the total number of registered snowmobiles in that subregion. 48 wide variety of information about each of the units in a survey; (2) many of the data items collected are classifications rather than continuous variables; (3) the difficulty in estimating the errors which are usually present in all the measures, not just the dependent variable; (4) complex probability samples, which pre­ sented problems in applying statistical techniques; (5) that among many of the explanatory variables used in the analysis, intercorrelations have existed that have made it difficult to access the relative importance of the different factors; and (6) the problem of handling the effects of interaction among the independent variables. They claimed it was a mistake to assume that their various influences on the dependent variable were addi­ tive, for two reasons. First, there were many known instances of strong interactions effects, where the effect of one variable on the dependent variable was dependent upon the value or presence of one or more other variables. Second, measured classifications were often proxies for more complex constructs. For example, age, marital status, number and age distribution of chil­ dren were all components of a life cycle construct.^ James N. Morgan and John A. Sonquist, "Problems in the Analysis of Survey Data, and a Proposal," Journal of the American Statistical Association, 58 (June 1963): 315-16. 49 In an attempt to improve the methods of analyzing survey data, Morgan and Sonquist developed a new technique for predicting^- social behavior from personal charac­ teristics called the Automatic Interaction Detector technique. 2 (AID) According to Sonquist it is: . . . a step-wise application of a one-way analysis of variance model. Its objective is to partition the sample into a series of non-overlapping sub­ groups whose means explain more of the variation in the dependent variable than any other such set of sub-groups.3 Analysis procedures.— The organization of the input data from this study and the operation of the algorithm used by the AID program were as follows. The data from each region were entered and analyzed sep­ arately. In each case the dependent variable was the number of snowmobiling days Figure 1). (reported in question 3, see The distribution of this variable was markedly skewed to the right. Sonquist and Morgan suggest that the extreme cases should be removed, The term "predictor" as used in this discussion only pertains to the amount of variation in the dependent variable that is explained by the independent variable and not in terms of a functional relationship. 2 John A. Sonquist and James N. Morgan, The Detection of Interaction Effects (Ann Arbor: Institute for Social Research, University of Michigan, 1964), pp. 180-217. 3 John A. Sonquist, Multivariate Model Building (Ann Arbor: Institute for Social Research, University of Michigan, 1970), p. 20. 50 therefore this skewness was reduced by deleting from the input data all values of 130 days or more.1 This procedure reduced the sample size in each region by approximately 10 percent. Weights, which summed to 1,000, were established for each region to take into account the different sampling fractions used (see Appendix J ) . The AID program collapsed the classes of the six­ teen independent variables into that dichotomous grouping which produced the greatest reduction in error variance in the sum of squares. Nominal variables were treated as free predictors, that is, the classes were monotonically reordered according to the mean values of the dependent variable. With ordinal or interval type var­ iables, however, the collapsing procedure required that the original order among classes be maintained. The optimal dichotomization on the independent variables was then treated as the "split" on that variable. This resorting process, of the "free" predictors, was repeated after each split. The total number of observations from each region treated separately was considered as the first parent group to be partitioned. squares The between-classes sum of CBSS^) was examined across the reordered 1Sonquist and Morgan, Detection of Interaction Effects, p. 120. 51 predictor variable. A cutting point was placed between contiguous groups of classes, for example, a versus b and c or a and b versus c, and the BSS calculated for p each dichotomous grouping. That grouping which contained the largest BSS^ indicated where the split would be made on that predictor, if it was chosen. This split took place between the two contiguous classes (after reorder­ ing) that had the largest BSS such that the parent group P was reorganized according to the resulting dichotomous class grouping of that predictor. In other words the two created groups contained c-r classes and r classes of the predictor. In the next and succeeding steps, all created unsplit groups were examined and the group i containing the largest total sum of squares mean was identified. for splitting, of: (TSS^) around its own This group i was then considered if it satisfied the eligibility criteria (1) possessing a TSS^ that was equal to or exceeded a specified minimum proportion (.015) of the total sum of squares for all input observations (TSST ) and contained a minimum number of observations (2) (50), to provide some sampling stability. The group thus selected became the new parent group. Each predictor variable was tested, over group i, to locate the predictor j_ that contained the largest BSS^ between two contiguous classes. Group i was then split 52 into two new groups, if the BSS satisfied a reducibility 1? criterion requiring that it be equal to or greater than the specified proportion of the TSST .^ This reducibility criterion was set in relationship to the size of the samples, which resulted in the following proportions: (1) .010 for Region I, (2) .009 for Region II, and (3) .006 for Region III. The above iteration process was terminated when one of the following conditions occurred: (1) no group contained a large enough TSS^ to satisfy that eligibility criterion; (2) no group contained 50 or more observations; (3) the reducibility criterion was not satisfied or; (4) the maximum number of unsplit groups allowable under the program limitations (50) was reached. Of the above con­ ditions, only the second and third were actually invoked in this application to terminate the iteration process; the other two remained, however, as safeguards. 2 The purpose in using the AID program was to identify the interactions that existed between the selected independent variables and not to develop a functional relationship, for as Herrmann states: ^The reducibility criteria were not discussed for the initial parent group as Sonquist and Morgan implied that it was not relevant to the first split (see Appen­ dix I) . 2 Sonquist and Morgan, Detection of Interaction Effects, pp. 5-6 and 158-61. These descriptions of the algorithm were placed in Appendix I . 53 The program does not, however, provide any estimates of the functional relationships between the pre­ dictor variables and the dependent variable. The technique is a useful preliminary to regression analysis but is not a substitute for it.l It was also of interest to determine the relative useful­ ness of each predictor variable in explaining the variation in the dependent variable. Robert 0. Herrmann, "Interaction Effects and the Analysis of Household Food Expenditures," Journal of Farm E conomics, 49 (November 1967): 831. CHAPTER III RESULTS AND ANALYSIS Phase I The purpose of Phase I was to determine and com­ pare, on a regional basis, the following: economic characteristics, (1) socio­ (2) ownership characteristics, (3) attitudes toward snowmobile regulations, and (4) patterns of snowmobile use. Contingency tables were developed which comprised the frequencies and percentages for each variable, each of the three regions. for Where there were appropriate population parameters available, these were included in the tables.'*' In some cases, statistics from two other snowmobile studies were included in the tables to facilitate comparisons. 2 ^Bureau of the Census, 1970 Census, Michi g a n . 2 Peter Klopchic, An Analysis of Snowmobiling m Ontario, Winter 1969-1970 (Toronto: The Department of Tourism and Information, 1971); Directional Marketing Company, 1970 Snowmobiler Survey (Duluth: Upper Great Lakes Regional Commission, 1971). 55 Wherever it was appropriate, tests of significance were applied. In the case of the quantitative variables, the Scheffe comparison of means was applied. An overall chi-square was calculated for the qualitative variables and then partioned by region. Socio-economic Characteristics Age characteristics.— The results of this study indicated that heads of snowmobile-owning households in Region I (the Upper Peninsula) were significantly older (statistically) than those from Regions II and III.. The mean age in Region I was 44.4 years compared to 42.6 years and 41.7 years in Regions II and III respec­ tively (see Table 3). There appeared to be a similar pattern in the age ranges of the general population. Compared to the percentage of those 20-24 years in the general population, the proportion of heads of snowmobile-owning households who were 25 years and under was much smaller. In the 25-34 year age group the pro­ portion of respondent household heads and persons in the general population were similar. However, 59 percent of the heads of snowmobile-owning households compared to 34 percent of the general population were between 35 and 54 years of age. There was little difference between the proportion of heads of snowmobile-owning households and members of the general population that 56 TABLE 3 AGE OF THE HEAD OF SNOWMOBILE-OWNING HOUSEHOLDS Region I Region II Percent Percent Age Region III No. No. State Percent No. SMO GPa 13 2 13 40 4 12 79 4 16 132 4 15 5 25 - 34 118 18 16 240 23 19 422 24 20 780 22 19 22 35 - 44 205 31 16 311 30 17 580 32 21 1096 31 21 31 45 - 54 199 30 18 293 28 18 469 26 12 961 28 13 28 55 - 64 105 16 18 133 13 16 207 12 16 445 13 16 21 3 19 28 3 18 31 2 16 80 2 16 661 100 100 1045 101 100 1788 100 101 3494 100 100 Under 25^ SMO GP Percent No. SMO GP SMO GP DMS ] 15 65 & over Total Mean 44.4 42.6 41.7 42. 5 SD 10.9 11.1 10.7 10. 9 CT * riT Key: Region I differed from Region II and from Region III (.05 level). SMO - Heads of "snowmobile-owning" households. GP - General population of region DMS - Directional Marketing Survey statistics aCalculated from: Bureau of the Census, 1970 Census, Michigan, Table 35. Age by Race and Sex for Counties: 197ETT ^For the general population, the number of 20-24 year olds was used to provide a comparative estimate. Calculation of percentages was based on the general population 20 years and over. 101 57 were 55-64 years of age. Only a small percentage of the general population in the 65 years and over category owned snowmobiles. In the Directional Marketing Survey (DMS), age distributions were almost identical to the results of this study.^ However, in Ontario, Klopchic estimated the average age of Ontario snowmobilers to be 38 years old 2 as compared to 42.5 years in this study. Education characteristics.— The average education level attained by respondent household heads in this study was 12.1 years, with those in Region III reaching a higher level than those in the other two regions (see Table 4). Heads of snowmobile-owning households with less than nine years of education comprised a smaller pro­ portion of the respondents to this study (particularly in Region I I ) , than in the general population. Respondent household heads in Region I who had completed some high school education appeared in the sample in the same pro­ portion as in the general population; in Region II and III the proportion in the sample was lower than that of ■^Directional Marketing Survey, p. 58. DMS will be used to refer to this survey. Hereafter 2 Klopchic, Snowmobiling in Ontario, p. 8. 58 TABLE 4 EDUCATION LEVEL OF HEAD OF SNOWMOBILE-OWNING HOUSEHOLDS Years of Education Completed Region I Region II Percent Percent No. GPa 84 13 17 9-11 128 20 12 248 38 126 20 1 - 4 Univ. ] Graduate 60 9 Total 646 100 Mean SD Comparison of Means Key: SMO GP 139 14 23 20 161 16 56 422 42 235 23 Percent 58 6 1016 101 Percent No. SMO GP 169 10 16 392 12 17 20 25 317 18 28 606 18 28 40 41 663 38 43 1333 39 43 20 482 27 843 25 ] 10 7 100 State No. No. SMO 8 or Less Region III 99 SMO GP 17 ) 12 126 7 1758 100 99 12 244 7 3420 101 12.0 11.9 12.2 12.3 2.7 2.5 2.5 2 .1 3 100 Region III differed from Region I and from Region II (.05 level) SMO - Heads of "snowmobile-owning" GP - General population of region ONT - Ontario snowmobilers households. Bureau of the Census, 1970 Census, Michigan, Table 120. and Family Characteristics for Counties: 3T91 W . bKlopchic, Snowmobiling in Ontario, p. 11. ONTb Educational 100 59 the general population. Those that completed high school made up the largest category in all regions. In Region II, 4 2 percent of the heads of snowmobile-owning households had completed grade twelve, which was similar to the per­ centage in the general population; however, in Region III, and particularly in Region I , the proportions were much less than in the general population of those regions. The proportion of respondent household heads that went beyond high school was three times as great in one sample from Regions II and III as it was in the general population; in Region I the proportion was four times as large. Klopchic's data show that snowmobilers in Ontario were not as well educated as their Michigan counterparts. Only 40 percent had completed high school, or better, compared to 71 percent of the snowmobilers in Michigan. Occupational characteristics.— The occupational characteristics of the heads of snowmobile-owning house­ holds varied between regions and also differed from the working population in the state (see Table 5). For Regions I and III, professional people were represented in the sample in similar proportion to that in the general population. In Region II, the percentage of "'"Ibid., p. 11. In comparisons between these two educational systems, it should have been noted that in Ontario, grade thirteen was required for high school completion. 60 TABLE 5 OCCUPATIONS OF HEAD OF SNOWMOBILE-OWNING HOUSEHOLDS Region II Region I Occupation Percent Region III Percent No. No. 95 9 13 200 12 12 8 168 16 7 306 18 6 20 76 7 20 162 140 22 15 270 26 15 Semi-Skilled 152 23 19 192 19 Cler/Sales Skilled 84 13 13 108 17 42 Percent No. GP Self-Emp/ Manager GPa Percent No. SMO Profession SMO State SMO GP SMO GP DMS 379 11 12 9 6 582 17 6 20 9 20 280 8 20 9 457 26 32 867 25 31 30 22 400 22 16 744 22 17 9 11 Service 60 9 17 72 7 15 66 4 10 198 6 Unskilled 10 2 6 21 2 6 23 1 4 54 2 4 ] 10 Farm Oper. 12 2 2 60 6 2 78 5 1 150 4 1 8 3 1 - 4 - - 7 - - 14 - - - 32 5 - 69 7 - 43 3 - 144 4 - 6 4 1 - 1 - - 2 - - 7 - - - 647 101 100 1029 99 100 1744 100 101 3419 99 102 101 Unemp/Stu.*3 Retired*3 Housewife*3 Total Key: SMO GP DMS - Heads of "snowmobile-owning" - General population of region - Directional Marketing Survey households aBureau of the Census, 1970 Census, Michigan, Table 122, Occupation and Earnings for Counties: 197UT ^These classifications were not included in Table 122. 61 professionals that were heads of snowmobile-owning house­ holds was less than the percentage of professionals in the general population. There was little difference between the percentage of respondent household heads in each region who were classified as self-employed or managers. However, repre­ sentation of this group in the sample for each region was two to three times as great as it was in the general population. Similarly, there was not much difference between the percentage of heads of snowmobile-owning households who were clerical or sales persons in each region. How­ ever, in the sample, clerical and sales personnel attained only one-third to one-half the representation that they actually had in the general work force. Even though the proportion of skilled workers in the general population of Region III was twice as great as in the other regions, the distribution in the sample was very similar for each region. In other words, a smaller proportion of skilled workers in the general population of Region III were heads of snowmobile-owning households than in the other two regions. The proportion of semi-skilled workers who were heads of snowmobile-owning households was greatest in Region I and least in Region II. general population, By comparison to the the percentage of semi-skilled 62 workers who were respondent household heads in Regions I and III was greater in the sample than in the general population, while in Region II the percentage was a little less. For service and unskilled workers the proportion found in the sample as heads of snowmobile-owning house­ holds was much less in Region III than in the other two regions. In all regions both groups were underrepresented by one-quarter to one-half in the sample, compared to the percentage of these workers in the general population. In Region I, farm households owning snowmobiles appeared in the sample in the same proportions as they did in the general population. However, for the other regions their representation in this survey was much higher. For example, snowmobile ownership by farm house­ holds appeared to be five times greater in Region III than their general population numbers would indicate. It should be noted that since land owners were not required to register their machine if it was only operated on the owner's property, then it was quite possible that farm operators who were heads of snowmobileowning households were not as accurately represented in this study as were persons with other occupations. Therefore, the percentage of farm households who owned snowmobiles could be much greater than this study indi­ cated. 63 The DMS found similar distributions according to occupation, with two exceptions. The proportion of semi­ skilled workers in their sample was less than one-half as large as in this study, while the percentage of farm operators was twice as large. In the DMS, the small sample of 500 could have resulted in a greater error in its estimates, especially when broken down into the eight occupational categories. Also, the basis on which respondents were assigned to the categories, skilled and semi-skilled, may have been different than the one used in this study.^ Income characteristics.— The mean incomes for snowmobile-owning households in Regions I, II, and III were estimated to be $10,900, $12,200, and $15,200 respectively significant. (see Table 6). These differences were The DMS and Klopchic study obtained state-wide and province-wide means of $10,700 and $10,900 respectively, compared to the overall mean of 2 $13,500 in this study. This discrepancy could have been due to differences in data collection techniques, as both studies established their highest category as $15,000 and over. In this study there were four ^"Directional Marketing Survey, p. 57. 2 p. 10. Ibid., p. 28; Klopchic, Snowmobiling in Ontario, 64 TABLE 6 COMBINED GROSS INCOME OF SNOWMOBILE-OWNING HOUSEHOLDS Region I Income3 Region II Percent Region III Percent No. Percent No. No. SMO GPb 77 13 35 91 9 29 $6,000 $9,999 261 42 35 333 35 $10,000 $14,999 186 30 22 315 $15,000 $24,999 74 12 8 $25,000 & over 16 3 614 100 Under $6,000 State SMO Percent No. GP SMO GP 46 2 19 214 7 20 31 342 20 21 936 29 23 33 27 614 37 29 1115 35 29 159 17 11 448 28 24 681 21 23 1 49 6 3 194 12 6 259 8 6 101 947 100 101 1644 100 99 3205 100 101 SMO DMS ] 44 35 1 21 Total Mean 4.2 4.6 5.4 5 .0 SD 1.5 1.6 1.6 1 .7 <"of^Means1' re9i°ns differed from each other Mean in Dollars0 $10,900 Key: GP $ 12,200 (.05 level) $15,200 $13,500 SMO - "Snowmobile-owning" households GP - General population of region DMS - Directional Marketing Survey aFor a description of the nine income categories see Figure 1. Bureau of the Census, 1970 Census, Michi g a n , Table 124, Income and Poverty Status in 1969 for C o u n t i e s : T 51W . These means were calculated using the mid-point of each category and using $3,000 and $30,000 to represent the lowest and highest category respec­ tively. 100 65 categories over $14,999 and the highest category was "over $30,000" so it would reflect the higher income groups with greater precision.'*' There was a distinct decrease from Region I to III in the percentage of households in the general popu­ lation who earned under $10,000. This pattern was even more pronounced with snowmobile-owning households; it decreased from 56 percent in Region I to 22 percent in Region III. This indicated that it was probably more difficult, or less attractive, for households in Region III with incomes less than $10,000 to become involved in snowmobiling than it was in the other two regions. The DMS obtained a proportion of 44 percent for this category compared to the 36 percent estimated for the state in 2 this study. Approximately a third of the snowmobile-owning households in the state earned between $10,000 and $14,999, which was the same proportion as that obtained by the DMS. ■*"The technique of collecting income data using categories may have had the advantage of obtaining a higher response, but it did present difficulties in calcu­ lating the means. In this study the categories were not in equal intervals (see Figure 1), therefore the means obtained by assigning the number 1 to 9 to each category, which assumed equal appearing intervals, produced relatively lower means than those resulting by using the dollar mid­ point of each category, which assumed that mean of each interval coincides with the mid-point. 2 Directional Marketing Survey, p. 8. 3Ib i d . 66 This proportion was 6 percent higher than that of the general population. However, this pattern was not con­ stant for each of the regions with the percentage of snowmobilers in that income group, increasing from 30 per­ cent in Region I to 37 percent in Region III. Snowmobile-owning households in the higher income brackets also comprised a larger proportion of the sample population (particularly in Region III) than households in these groups did in the general population; the pro­ portion in these higher groups ranged from 15 percent in Region I to 40 percent in Region III. However, the DMS estimated that there were 10 percent fewer snowmobilers in this category. The small sample that the DMS used or the fewer high income classes used may have contributed to the differences in the estimates obtained.'*' Number of children.— The differences between regions in the mean family size appeared to be small Table 7). (see The DMS estimate of 2.2 children per family was much higher than the 1.8 calculated in this study. However, the mean calculated by the DMS included children 19 years to 21 years old, which would not only have con­ tributed to a larger mean but also a smaller percentage The DMS used a registration listing that did not include some 60,000 snowmobile owners who registered between June 1969 and April 1970. TABLE 7 NUMBER OF CHILDREN 18 AND UNDER IN EACH SNOWMOBILE-OWNING HOUSEHOLD N o . of Children None One Two Three Four Five + Total Region I g, o No. Region II g. o No. Region Ill Q. O No. State o o No. DMSa ONTb 181 144 139 81 55 53 28 22 21 12 8 8 269 183 233 154 107 90 26 18 22 15 10 9 549 304 386 284 180 75 31 17 22 16 10 4 999 631 758 519 342 218 29 18 22 15 10 6 18 19 23 18 13 9 16 33 26 12 13 653 99 1036 100 1778 100 3467 100 100 100 Mean 1. 8 1. 9 1. 7 1. 8 SD 1. 7 1. 7 1. 6 1. 6 Region II differed from Region I and from Region III (2.2) (1.9) (.05 level). -a Results of the Directional Marketing Survey. children up to and including 21 years of age. These figures included Results obtained by Klopchic in Ontario. Since the number of families without children was not reported the percentages given for the number of children per family provide only rough comparisons. 68 of owners who were classified as having no children. In fact, if these two means had been calculated without including the zero children category when the means would have been almost identical. Klopchic produced an estimate of 1.9 children per family in Ontario. 2 The relatively small percentage of owners in Region II who did not have children under 19 years of age may in part been due to the high proportion of retired snowmobilers in this region (see Table 5) . In Region III only half the proportion of owners had fami­ lies of five or more children, compared to the other regions. Membership in snowmobile clubs.— Only one-eighth of the total sample joined snowmobile clubs (see Table 8). However, respondents in Region I became club members to a greater extent than those from the more urban regions. Snowmobiling groups preferred.— There appeared to be little regional difference in the type of groups with whom snowmobile owners preferred to snowmobile, with the exception of "friends" and "family" (see Table 9). In Region I, friends were preferred for companions to a ■*"Ibid. , p. 21. 2 Klopchic, Snowmobiling in Ontario, p. 9. 69 TABLE 8 SNOWMOBILE CLUB MEMBERSHIP BY SNOWMOBILE-OWNING HOUSEHOLDS Region I Region II No. % No. Yes 124 20 122 12 187 11 433 13 No 511 81 833 88 1515 89 2909 87 635 101 1005 100 1702 100 3342 100 Region III State Membership Total Source o. o No. df % X No. 2 % P Partition of Chi-square I vs II & III 1 27.43 .001 II vs III 1 .82 NS 2 28.25 .001 Total 70 TABLE 9 TYPE OF SNOWMOBILING GROUPS PREFERRED Region I Region II Region III No. No. No. State Groupa % % No. % % Alone 43 7 54 5 71 4 168 5 Wife 54 8 94 9 149 8 297 9 Children 39 6 83 8 147 8 269 8 300 46 398 38 609 34 1301 37 Org. Group 13 2 21 2 44 3 78 2 Family 52 8 128 12 305 17 495 14 7 1 7 1 16 1 30 1 148 23 265 25 451 25 864 25 656 101 1044 100 1792 100 3492 101 Friends Girl Friend/ Boy Friend Other Total ci These groups were not mutually exclusive. Respondents were asked to indicate the group with which they snowmobiled most of the time. 71 greater extent than in the other regions. In the case of preference for family groups, the pattern was reversed. When these groups were combined, the regional differences tended to disappear. Ownership Characteristics Snowmobile horsepower.— There was a relationship between the region in which a snowmobiler lived and the horsepower of the machine he owned (see Table 10). As one went from Region I to Region III the machines became increasingly more powerful. Klopchic estimated that snowmobiles in Ontario were slightly less powerful: 18 horsepower compared to 20 horsepower in Michigan. Years of snowmobiling.— The year in which present snowmobile owners bought their first machine was related to their region of residence (see Table 11). In other words, snowmobiling in Michigan tended to start in the north, in Region I, and gradually worked its way down to Region III. A larger proportion, 14 percent, of the sample in Region I bought their first snowmobile in the mid 1960's compared to only 3 percent in Region III. However, by the spring of 1970 this situation had reversed itself. K l opchic's data indicated that the mean purchase date in Ontario was August 1968 compared to January 1968 in this study.^ One-third of the respondents reported ^"Ibid. , p . 39 . 72 TABLE 10 SNOWMOBILE HORSEPOWER3 Region I Region II Region Ill No. % No. No. 1-17 265 43 326 32 18 - 22 185 30 340 23 + 170 27 620 100 State Horsepower Total Mean SD Comparison of Means % No. % 419 25 1021 30 34 605 36 1135 34 337 34 677 40 1199 36 1005 100 1701 101 3356 100 % 18 .8 20.0 21. 0 20. 3 6 .7 6.3 6. 4 6. 5 All regions differed from each other level) (.05 The data were based on the first snowmobile reported. The means and SD were calculated before the data were grouped. 73 TABLE 11 YEAR OF FIRST SNOWMOBILE PURCHASE Year of First Purchase Region I No. o. o Region II Region III No. No. % State No. % % Before 1966 88 14 70 7 63 3 221 7 1966 17 13 72 7 79 5 230 7 1967 125 20 172 17 280 16 577 17 1968 133 21 259 26 450 26 842 25 1969 164 26 342 34 645 38 1151 34 1970 37 6 85 9 199 12 321 10 626 100 1000 100 1716 100 3342 100 Total Mean SD Comparison of Means 1967 .4 1967 .9 1968 .2 196 8 .0 1 .7 1 .4 1 .3 1 .4 All regions differed from each other level) (.05 74 that they made their first purchase in 1969 which empha­ sized the rapidly expanding nature of this outdoor activity. Number of snowmobiles per household.— There appeared to be no particular relationship between the region in which a snowmobiler resided and the number of snowmobiles owned by that household (see Table 12). The mean number of machines owned was 1.4, which was very close to 1.4 5 machines owned that the DMS estimated.'*' In Ontario, Klopchic estimated that the average snowmobile household owned 1.2 machines, which may have reflected the 2 lower mean income of snowmobilers in that Province. Other equipment owned by households.— Snowmobilers generally appeared to own a substantial number of motor­ ized recreation vehicles (see Table 13). Twenty-four percent of the respondents owned truck campers or house trailers, with the regional proportions gradually increasing from 18 percent in Region I to 27 percent in Region III. Ownership of motorcycles followed the same pattern having almost identical proportions in each region. Snowmobile trailers (to transport snowmobiles) were owned by 60 percent of the respondents in the state. "^Directional Marketing Survey, p. 16. 2 Klopchic, Snowmobiling in Ontario, p. 35. 75 TABLE 12 NUMBER OF SNOWMOBILES PER HOUSEHOLD Region I Region II Region Ill State Number NO. O, o NO. o o No. % No. % One 624 72 1006 71 1707 73 3337 72 Two 204 24 333 24 550 23 1087 24 30 4 62 4 72 3 164 4 4 1 15 1 17 1 36 1 862 101 1416 100 2346 100 4624 101 Three Four Total Mean 1. 4 1 .4 1. 4 1. 4 SD 0. 65 0 .65 0. 60 0. 61 Comparison of Means The F was not significant at the .05 level, 76 TABLE 13 OWNERSHIP OF OTHER VEHICLES BY SNOWMOBILE-OWNING HOUSEHOLDS a Region I Region II Region III No. % No. No. 125 19 249 24 474 26 848 24 20 3 27 3 52 3 99 3 5 1 18 2 22 1 45 1 Power Boat 240 36 360 34 664 37 1264 36 SM Trailer 369 56 522 49 1221 68 2112 60 House Trailer 68 10 110 10 243 14 421 12 Truck Camper 55 8 120 11 234 13 409 12 882 133 1406 133 2910 162 5198 148 Type of Vehicle Motorcycle All Terrain Vehicle SM Conv. Kit Total Number of Respondents 665 1060 % 1802 State No. % % 3527 Respondents were asked to check all equipment owned by the household, which resulted in multiple count­ ing. 77 The regional proportions decreased from 56 percent in Region I to 49 percent in Region II and then increased to 68 percent in Region III. These figures did not necessarily provide a good estimate of the owners' capability to transport snowmobiles, as they did not take into account the possibility of trailer sharing by owners (many trailers can carry two machines), nor the number of respondents who had "pick-up" trucks which could be used for this purpose. The overall proportion of respondents who owned power boats was 36 percent; for each region. this pattern was very similar However, in the general population of the state the percentage of households that owned power boats was only 16 percent. In Regions I and II, the per­ centage of households in the general population which owned power boats was 29 percent and 31 percent, respec­ tively, whereas in Region III it was only 14 percent.^ It appeared that households in the sample from the two northern regions owned power boats to slightly greater degree than those in the general population. However, ownership by snowmobilers in Region III was two and a half times as great as in the general population. ■*"The number of registered power boats (1971) was obtained from the Michigan Secretary of State's Office. These were: Region I, 30,670; Region II, 68,850, and Region III, 368,350. 2 Bureau of the Census, 1970 Census, Michigan, Table 124 (Income and Poverty Status m 1969 for 78 Even though the above data indicate that snow­ mobilers tended to have a greater amount of motorized recreation vehicles than did the general population, it may have been more closely related to income than to snowmobile ownership. Opinions About Selected Regulations Present regulations.— The majority (60%) of all respondents felt that the present regulations should remain unchanged. Twenty-nine percent were of the opinion that they should be made stricter while 12 per­ cent held the opposing view (see Table 14). There was a relationship between the opinion held and the region of residence. Snowmobile owners in Region II were more in favor of making the present regulations more strict than were the respondents from Regions I or III. Enforcement of regulations.— Though only 29 per­ cent of all respondents felt that the present regulations should be made more restrictive (see Table 14), 44 percent were of the opinion that the enforcement of them should Counties.) This table contained the number of families and unrelated individuals by county, excluding inmates of institutions, members of the armed services, college students in dormitories, and unrelated individuals under 14 years. TABLE 14 SHOULD THE PRESENT SNOWMOBILING REGULATIONS BE MORE STRICT Region I Opinion Region II Region III State Code NO. o. o O. o No. No. g. o No. % Much more strict 1 41 6 49 5 62 4 152 5 More strict 2 145 22 268 26 392 22 805 24 Unchanged 3 365 56 591 58 1067 62 2023 60 Less strict 4 89 14 100 10 179 10 368 11 Much less strict 5 8 1 8 1 29 2 45 1 648 99 1016 100 1729 100 3393 101 df X I & II vs III 4 29.96 .001 I vs II 4 10.42 .05 Total 8 37.38 .001 Total Source 2 P Partition of Chi-square 80 be stricter (see Table 15). Only 9 percent felt the enforcement of regulations should be relaxed. The data indicated that there was an association between the region of residence and the opinion expressed. Region II respondents were more in favor of stricter enforcement practices. Operation of snowmobiles by those 14 years and under.— The majority (77%) of the respondents were opposed to allowing children 14 years old and under to operate snowmobiles without adult supervision. a large proportion However, (85%) were in favor of permitting this age group to operate a machine if it was done within view of a supervising adult (see Table 16). No association was detected between the opinions expressed by the respondents and their region of residence. Regulation of the noise level.— The overall response indicated that two-thirds of the snowmobile owners were in favor of regulating the amount of noise a snowmobile should be allowed to emit (see Table 17). However, the respondents from Region III differed in their expressed opinion from those in the other two regions, being less in favor of noise controls. Regulation of snowmobiling near ice-fishermen.— Three-fifths of all snowmobile owners in the sample were in favor of establishing regulations that would control TABLE 15 SHOULD ENFORCEMENT OF SNOWMOBILING REGULATIONS BE MORE STRICT Opinion Region I Region II Region III No No No State Code Much more strict 43 79 More strict 235 36 413 Unchanged 290 45 75 12 Less strict 647 Source 101 40 600 35 1248 37 450 872 50 1612 47 73 129 277 20 30 1023 100 1731 df Partition of 16.82 01 I vs II 11.04 05 Total 27.86 001 Chi-square 232 110 Much less strict Total No 100 3399 100 82 TABLE 16 SHOULD CHILDREN 14 YEARS OF AGE AND UNDER OPERATE SNOWMOBILES Region I Region II Region III No. No. No. State Opinion % % % No. % Without Supervision Yes 118 23 205 25 295 21 618 23 NO 388 77 612 75 1089 79 2089 77 506 100 817 100 1384 100 2707 100 Total Chi-square 2 The total X 2 = 4.24, which was not significant at the .05 level. With Supervision Yes No Total Chi-square 419 86 746 85 1315 85 2552 85 84 14 136 15 326 15 456 15 575 100 The total 100 100 1551 3008 882 100 2 = 1*59, which was not significant at the .05 level. 83 TABLE 17 SHOULD SNOWMOBILE NOISE LEVELS BE REGULATED Region I Region II Region III NO. No. NO, State Opinion NO. Yes 444 69 721 71 1135 65 2300 67 No 200 31 300 29 619 35 1119 32 644 100 1021 100 1754 100 3419 99 Total Source df x2 P I & II vs III 1 10.74 .001 I vs II 1 .50 NS 2 11.24 Partition of Chi-square Total .01 84 snowmobile activity near persons engaged in ice-fishing (see Table 18). However, owners in Region I were dif­ ferent in their attitude from those in the other two regions, being about evenly divided on this issue. Allowing snowmobiling on public thoroughfares.— A large majority (93%) of the respondents were opposed to permitting snowmobiling along main highways Table 19). (see No regional differences were found in the respondents' opinion on this question. However, in responding to the question about allowing such activity on secondary highways, snowmobile owners across the state were evenly divided in their opinion (see Table 19). A significant difference was found between the opinions of the respondents from Region III and those from the two other regions. Those from Regions I and II were less in favor of permitting such activity. A large majority (89%) of all respondents favored allowing snowmobiling along highway shoulders (see Table 19). Snowmobilers in Region III were somewhat less in favor. In response to the question about permitting snowmobiling on town and village streets, three-fifths were in favor of allowing such practices (see Table 19). However, the opinions expressed varied more from region to region than on any other question (although the 85 TABLE 18 SHOULD SNOWMOBILING NEAR PERSONS ICE-FISHING BE REGULATED Region I Region II Region III No. % No. % No Yes 335 53 602 60 NO 303 48 402 638 101 1004 State Opinion Total Source Q. •Q NO. 1062 62 1999 60 40 657 38 1362 41 100 1719 100 df . x2 % 3361 101 P Partition of Chi-square I vs II & III 1 15 .87 .001 II vs III 1 .87 NS 2 16 .74 .001 Total 86 TABLE 19 SHOULD SNOWMOBILES BE ALLOWED TO DRIVE ON A: Region I Region II Region III State Opinion No. % No. % No. % No. % Main Highway Yes No Total 36 552 6 94 67 875 7 93 128 1467 8 92 231 2894 7 93 588 100 942 100 1595 100 3125 100 .--------- ---------- — 5------- ---------- ------ Chi-square The total ~ 2.29, which was not sig­ nificant at the .05 level. Secondary Highway Yes No Total 261 340 43 57 462 504 48 52 882 768 53 47 1605 1612 50 50 601 100 966 100 1650 100 3217 100 Source Partition of Chi-square I & II vs III I vs II Total 2 df X 1 1 17.20 2.87 .001 NS 2 20.07 .001 P Highway Shoulder Yes No Total 587 64 90 10 929 88 91 9 1517 226 87 13 3033 378 89 11 651 100 1017 100 1743 100 3411 100 Source Partition of Chi-square I & II vs III I vs II Total df x2 P 1 1 12.84 .56 .001 NS 2 13.40 .01 87 TABLE 19— Continued Region I Region II Region III State Opinion No. % No. % No. % No. % Town or Village Street Yes No Total 170 429 28 72 324 616 35 66 768 838 48 52 1262 1883 40 60 599 100 940 101 1606 100 3145 100 df x2 1 1 80.85 5.64 .001 .05 2 86.47 .001 Source Partition of Chi-square I & II vs III I vs II Total P 88 partitioning did not test each p a r t ) . The more urban the region the more inclined those owners were to favor permitting snowmobiling on town and village streets.^ Patterns of Snowmobile Use Snowmobile owners were asked to indicate their participation in snowmobiling measured in number of days, by counting any part of a day spent on the sport as one day. For each of the following classes of questions they were asked to report the number of days spent: (1) snow­ mobiling in each of the three counties of most use, (2) snowmobiling on different classes of land, a particular snowmobiling activity, (3) doing (4) participating in activities associated with a snowmobile trip, and (5) on overnight trips. questions, except In each of the above classes of (1), double counting of days was expected to occur in varying degrees. Very little double counting was expected in the reporting of the number of days spent in the three counties of most use. However, for the other four classes of questions it was expected that the number of days reported for each class would exceed the actual number of days reported in Question 3. In other words, a snowmobiler may have used ^Snowmobilers from Region III did a large portion of their snowmobiling outside their county of residence (which will be discussed later), hence, it is possible that this was an expression, by some, favoring snowmobil­ ing on someone else's streets. 89 three classes of land in one day, so this would appear as three days in Question 6, but only be represented as one day in Question 3. The number of days reported ranged from zero to ninety-five d a y s .^ For most of the activities there was a large variation in the amount of use that was reported which was reflected by large standard deviations. Amount of snowmobile u s e .— When the total amount of snowmobile use was examined (Question 3), the results indicated that the respondents from Region III did less snowmobiling than those from each of the other two regions (see Table 20). No significant difference was observed between the respondents of Region I and of Region II. However, when the amount of snowmobiling done in the owners' county of residence was considered, the number of days reported by respondents differed according to region (see Table 20). The mean number of snowmobiling days reported for the county of residence ranged from twenty-seven days in Region III to forty-nine days in Region I. The lower average number of snowmobiling days reported by owners in Region III was probably closely related to the fact that Region III received much less ^"Where individuals reported a number in excess of ninety-five days, it was coded as ninety-six. The maximum number of respondents reporting over ninety-five days in any activity was 5 percent. TABLE 20 TOTAL AMOUNT OF SNOWMOBILE USE PER SNOWMOBILE OWNER DURING 1969-70 SEASON Total Number of Daysa Days in Resident County Region Days Owners Mean SD Days % of Tot. Mean SD Region I 33,500 596 56 35 29,400 88 49 30 Region II 51,400 949 54 35 41,800 82 44 28 Region III 69,300 1649 42 29 44,200 64 27 23 State 154,200 3194 48 32 114,000 74 35 28 Comparison of Means Region III differed from Region I and from Region II (.05 level). All regions differed from each other (.05 level). aThe total number of days spent snowmobiling was calculated by summing the number of days spent snowmobiling in the county of: (1) most use, (2) second most use, and (3) third most use. 91 snowfall than the other two regions (see Appendix E ) . Another factor which probably acted as a constraint was the relative lack of public lands available; none in the case of national and state forests (see Appendix H ) . The available lands also had to serve a much larger population. It was also likely that the above factors were closely associated with Region III snowmobilers using counties other than their own 36 percent of the time. Indications were that a considerable portion of this time was spent in counties outside of Region III. Approximately one-third of these respondents spent an average of seven days taking two- to three-day trips of 100 miles or more from their place of residence (see Table 24). Ownership of land used for snowmobiling.— It was expected that the relationships observed with respect to the total amount of time spent snowmobiling/ discussed above, would be reflected in the responses to the amount of time spent on different types of land. appeared to be the case. In general, this Respondents from Region III dif­ fered from those in the other two regions with respect to use of each classification of land except the category "other" where no differences were observed (see Table 21). Again, no differences were observed in the use patterns between snowmobile owners from Region I and Region II TABLE 21 AMOUNT OF SNOWMOBILING ON VARIOUS CLASSES OF LAND PER SNOWMOBILE OWNER Region I 1000 Days % of Usea Region II °* 1000 Days % of Usea Region III 1000 Days °* State % of Usea * 1000 Days % of Usea 30 962 65.7 29 XT Snowmobile Owner's Land 13.0 25 377 25.0 30 638 27.6 Mean 34 39 28 33 SD 31 31 24 28 Comparison of Means All regions differed from each other 1977 (.05 level). Private Land 12.2 23 399 18.9 23 633 24.8 27 1150 55.9 25 Mean 31 30 22 26 SD 28 27 20 24 Comparison of Means Region III differed from Region I and from Region II (.05 level.) 2182 TABLE 21— Continued Region I 1000 Days % of Usea Region II °* 1000 Days % of Usea Region III „ State 1000 Days % of Usea °* 1000 Days ^ Usea No 1 X10• 16.7 18 1065 42.0 19 2071 State Land 9.5 18 358 15.8 19 648 Mean 27 24 16 21 SD 26 24 16 21 Comparison of Means Region III differed from Region I and from Region II (.05 level) • Local Roads 5.5 11 269 10.3 13 446 6.8 7 579 23.5 10 Mean 20 23 12 17 SD 24 26 15 21 Region III differed from Region I and from Region II (.05 level). 1294 TABLE 21— Continued Region I 1000 Days % of Usea Region II M 1000 Days Region III % of Usea 1000 Days State % of Usea 1000 Days % of Usea County Land 5.9 11 215 5.2 6 219 6.2 7 362 17.3 8 Mean 27 24 16 21 SD 26 26 18 23 Comparison of Means 796 Region III differed from Region I and from Region II (.05 level). Federal Land 3.6 7 181 3.1 4 161 4.0 4 277 10.7 5 Mean 19 19 13 16 SD 21 22 14 19 c°m Parison Region III differed from Region I and from Region II (.05 level). 619 TABLE 21— Continued Region I 1000 r** Days % of tt a. Use Region III Region II No. 1000 Days % of Usea „ 1000 Days % of Usea State „ * 1000 Days 241 11.7 of Usea No, Other Land*3 2.5 5 89 3.9 5 149 5.3 6 5 Mean 28 26 22 24 SD 25 25 23 24 Comparison of Means 479 The F ratio was not significant at the .05 level. Total Usec 52.2 Mean 100 28 1888 82.2 100 28 2894 91.4 99 20 4636 226.8 101 9418 24 The percentage of total use on all classes of land reported by respondents. In checking the category "other," respondents from each region indicated that more than half of this time was spent snowmobiling on lakes and rivers. The large combined number of responses indicated the extent of double counting. 96 except in the use of the owners' own land. In the latter instance, those from Region II spent a greater number of days snowmobiling on their own land, compared to owners from each of the other two regions. It was interesting to examine the proportion of time respondents from each region spent using each of these types of land. Snowmobilers from all regions spent the same percentage of their time on state owned land. The proportion of time using local roads for snowmobiling varied from 7 percent in Region III, 11 percent in Region I to 13 percent in Region II. of the snowmobilers' A greater proportion time in Region I was spent using county owned land compared to those in the other two regions. days. Federal lands were used the fewest number of However, respondents from Region I spent almost twice as large a proportion of their time on this clas­ sification of land than did those from the rest of the state. Owners from Region II and III spent approximately the same percentage of their time using their own land, while those from Region I spent a smaller proportion. In Minnesota it was calculated that snowmobilers spent 49 percent of their time on private land compared to 47 percent on public land.'*' When the above categories (excluding "other") were collapsed, it was found that ^Bureau of Planning, Minnesota Snowmobile S t udy, 1970 (St. Paul, Minn.: Department of Conservation, 1970), p. 11. 97 snowmobilers in Michigan spent 54 percent of their time on privately owned land compared to 42 percent on public land. However, the proportions observed in Region I were almost identical to the results obtained in Minnesota. Kinds of snowmobile activity.— Snowmobile owners in Michigan spent a greater mean number of days trail riding than doing any other snowmobiling activity. Respondents from each region differed from those in each other region in this respect, spending an average of forty days in Region I, thirty-five days in Region II, and twenty-five days in Region III (see Table 22). Snow­ mobilers in Region III spent a smaller average number of days "scrambling” (that is, snowmobiling in open areas) differing considerably from those in Region I and in Region II. To an even greater degree, respondents in Region III differed from those in the other two regions in using the machines for travelling to work. However, no regional differences appeared in the amount of time spent racing snowmobiles. In examining the proportion of time spent in these activities, it was found that a greater percentage (58%) of snowmobiling was done on trails compared to scrambling (31%). However, respondents from Region III devoted a greater proportion of their time (35%) to scrambling compared to those in the other two regions (27% and 29%). In Minnesota it was estimated that TABLE 22 KINDS OF SNOWMOBILE ACTIVITY PER HOUSEHOLD Region II Region I 1000 Days % of Usea No* 1000 Days Region III % of Usea 1000 Days % of Usea 34.1 54 State 1000 Days % of Usea 82.8 58 Trail Riding 20.6 64 520 28.2 59 815 1376 Mean 40 35 25 31 SD 29 28 22 26 Comparison of Means All regions differ from each other 2711 (.05 level). Scrambling 9.5 29 355 12.8 27 486 22.2 35 1002 44.5 31 Mean 26 26 22 24 SD 25 25 20 23 Comparison of Means Region III differed from Region I and from Region II (.05 level). 1843 TABLE 22— Continued Region I 1000 Days % of Usea Region II N 1000 Days % of Use3 Region III M No' 1000 Days % of Use3 State 1000 Days No' % Of use3 »-T No- Riding to Work 1.6 5 66 2.2 5 132 1.3 2 125 5.1 4 Mean 24 17 10 16 SD 31 24 16 23 Comparison of Means Region III differed from Region I and from Region II 323 (.05 level). Snowmobile Racing 0.4 1 43 0.5 1 77 1.0 2 149 1.9 1 Mean 10 6 7 7 SD 14 7 11 10 ^°m Parison T^e F ratio was not significant at the .05 level. 269 TABLE 22— Continued Region I 1000 Days % of a. Usea TT Region II M No • 1000 Days % of Usea Region III 1000 Days N % of Usea State „ 1000 Days % of Usea N Other Activities 0.2 Mean 1 71 3.7 8 135 4.5 27 3 7 189 8.4 24 6 395 21 Totals 32.3 Mean 100 31 1055 47.4 100 29 1645 63.1 100 22 2841 142.7 100 5541 26 aThe percentage of the total amount of snowmobile activity reported for each household. 101 snowmobilers spent 62 percent of their time on trails compared to 38 percent in open areas.^ However, if only trail riding and scrambling were included in the calcu­ lations for Michigan, then the adjusted estimates, 60 per­ cent and 40 percent, were very similar to those obtained in Minnesota. In all regions using a snowmobile for travelling to work or participating in snowmobile races made up a small proportion of snowmobile activity respectively) mobilers. (4% and 1% and also involved relatively few snow­ In Minnesota, the pattern appeared to be similar as they estimated that 2 percent of the snow­ mobilers used their machine to travel to work and 1 per2 cent were involved m racing. However, from the above data it could not be determined whether snowmobilers preferred trail snow­ mobiling to open space snowmobiling or whether the nature of the available resources determined the patterns observed. It did appear that there could be a relation­ ship between the use of local roads (see Table 21) and using one's snowmobile for going to work. The proportions of time spent on both of these pursuits were much higher in Region I and II than in Region III. ^Ibid. , p . 13. 2I b i d ., p. 21. 102 Activities associated with snowmobiling.— These activities were considered to be associated with or secondary to the principal activity of snowmobiling; hence, it was expected that the means would be relatively low (see Table 23). A comparison of the mean number of days indicated that snowmobiling households in Region III differed from those in Regions I and II in spending less time ice-fishing and on cook-outs, but spent more time overnight camping. No regional differences were observed in the case of hunting. However, households in Region I differed from those in the other regions in spending more days per household tobogganing, sledding, and skiing. In examining the proportion of time spent on each of these activities by snowmobiling households, it was apparent that ice-fishing was the most popular activity associated with snowmobiling; households throughout the state spent approximately one-third of the time in this manner.'*" Having a cook-out while on a snowmobile outing appeared to be more common in Region II as they spent 26 percent of the time in this way compared to 17 percent and 18 percent in the other regions. time spent hunting The proportion of (14%) was similar in all regions. The calculation of these proportions was based on the total amount of time reported spent on these associated activities. In other words the percentage of time spent at one of these activities should not be identified with a percentage of total time spent snow­ mobiling . TABLE 23 ACTIVITIES UNDERTAKEN PER HOUSEHOLD IN ASSOCIATION WITH SNOWMOBILING Region II Region I 1000 _ Days % of M rT a No. Use0- 1000 _ Days % of a Use State Region III „T No • 1000 Days % of Usea 1000 Days % of a Use 11.1 32 „o. N Ice Fishing 2.9 33 196 3.9 35 316 4.3 30 454 Mean 15 17 9 11 SD 18 12 13 16 Comparison of Means 966 Region III differed from Region I and from Region II (.05 level). Cook-outs 1.6 Mean SD Comparison of Means 18 208 2.9 26 371 2.4 17 349 6.9 20 8 8 5 7 10 11 6 9 Region III differed from Region I and from Region II (.05 level). 928 TABLE 23— Continued Region I 1000 Days % of Use3 Region II „ No‘ 1000 Days Region III % of Usea 1000 Days State % of Usea 1000 Days % °f . . a Use No • Hunting 1.2 Mean SD Comparison of Means 14 125 1.5 13 165 2.2 15 283 4.9 14 9 9 8 8 10 12 8 10 573 The F ratio was not significant at the .05 level. Tobogganing, Sledding, and Skiing 1.0 11 84 1.1 10 142 2.2 15 272 4.3 13 Mean 12 7 8 9 SD 16 10 10 11 Comparison of Means Region I differed from Region II and from Region III (.05 level). 498 TABLE 23— Continued Region I 1000 Days % of Usea Region II w 1000 Days % of Usea Region III NO' 1000 Days % of Usea State M N° ‘ 1000 Days % 0 f 3. Usea TT XT — NO • Overnight Camping 0.5 6 86 0.3 3 58 1.3 9 127 2.2 6 Mean 6 6 10 8 SD 6 5 13 10 Comparison of Means 271 Region III differed from Region I and from Region II (.05 level). Other Activities 1.6 Mean 18 53 1.6 30 14 59 1.8 13 101 5.0 18 27 15 213 23 Totals 8.8 Mean 100 12 752 11.3 101 10 1111 14.2 99 9 1586 34.4 100 3449 10 aThe percentage of the total amount of activity associated with snowmobiling reported for each household. 106 Households in Region III spent a greater percentage of the time (15%) tobogganing, sledding, and skiing than did those in the other regions. Overnight snowmobile trips.— The response to question 16, which asked snowmobile owners to report the number of their overnight trips, resulted in a number of the fifteen cells in the five-by-three chart having very low frequencies. Therefore, the categories of both the time and distance variables were collapsed, as indicated in Table 24. No regional differences (significant at the .05 level) were observed in the number of overnight trips taken or in the distances travelled. However, it was of interest to examine the different proportions of trips taken in each region. Distance appeared to be a dominant factor for respondents in all regions. In Region I, 86 percent of the trips reported were less than 100 miles from the respondent's home. However, in Region III, 71 percent of the trips were over 100 miles from the owner's residence. In Region II, the influence of dis­ tance did not appear to be as strong, as 68 percent of the respondents travelled less than 100 miles from home, but 58 percent took trips extending over two or three nights. It appeared that respondents from Region III had a greater tendency to travel long distances than those from the other two regions, especially when it TABLE 24 OVERNIGHT SNOWMOBILE TRIPS Region I Trips %a Region III Region II No.b SM % No. Trips % State No. Trips % No. One--Night Trips Less Than 100 Miles 793 Mean SD Comparison of Means 57 140 660 34 111 1511 16 192 2964 23 6 6 8 7 12 8 13 11 443 The F ratio was not significant at the .05 level. One-Night Trips More Than 100 Miles 73 5 16 126 7 47 1933 20 314 2132 17 Mean 5 3 6 6 SD 3 3 10 9 ^°m Parbson T^e F ratio was not significant at the .05 level. 377 TABLE 24— Continued Region I Trips %a Region II No.*3 SM % Region III No. Trips % State No. Trips % No. Two- and Three-Night Trips Less Than 100 Miles 399 Mean SD Comparison of Means 29 62 660 34 79 1210 13 146 2269 18 6 8 8 8 13 15 14 14 282 The F ratio was not significant at the .05 level. Two- and Three-Night Trips More Than 100 Miles 130 Mean SD Comparison of Means 9 15 469 24 90 4913 51 707 5582 43 9 5 7 7 23 12 10 11 822 The F ratio was not significant at the .05 level. Total Trips 1395 100 233 1915 99 327 aPercentage of total number of trips. Id Number of snowmobile owners responding. 9567 100 1360 12947 101 1920 109 was noted that approximately one-fifth of those who reported trips in excess of 100 miles actually took trips in excess of 300 miles. Expanded Snowmobile Use Days It was decided to expand the data to obtain a better picture of the magnitude and distribution of snow­ mobile activity in Michigan. The calculation of the weighting factors was based on the proportion of the number of usable responses to the number of registered snowmobiles in each of the eleven sub-regions Appendix G ) . (see Using these weights the data were expanded for each county. In some counties, the sample was small (less than twenty) which increased the probability of sampling variation. Therefore, whatever errors occurred in the estimates were magnified in the expansion. How­ ever, when examined at a regional level the data were expected to provide good estimates of the snowmobiling activity. The expansion of the data was placed in Table 25. The first column contained the total number of snowmobile days generated by residents of each county. The dif­ ference between this number and the number of days spent snowmobiling in the county of residence resulted in the number of "out-flow" days; that is, the number of days spent snowmobiling outside one's county of residence. The number of days spent in each county by both residents TABLE 25 EXPANDED SNOWMOBILE USE DAYSa - IN THOUSANDS OF DAYS - County Tot Dys From Cty of Origin*3 Dys in Cty of Origin0 Out Flow Days Dys in Cty of Destin° In Flow Dc.ys Net Gain or Loss Percent Gained Or Lost Region I Alger Baraga Chippewa Delta Dickinson Gogebic Houghton Iron Keweenaw Luce Mackinac Marquette Menominee Ontonagon Schoolcraft 43 32 141 139 63 48 51 45 1 40 61 175 43 52 44 38 30 134 110 54 38 44 39 1 32 50 159 41 49 39 5 2 7 29 9 10 7 6 0 8 11 16 2 3 5 69 37 154 116 57 47 50 49 9 43 74 180 49 60 51 31 7 20 6 3 9 6 10 8 11 24 21 8 11 12 Total 978 858 120 1,045 187 + + + + .+ + + + + + + 26 5 13 23 6 1 1 4 8 3 13 5 6 8 7 + 65 + 61 + 16 + 9 - 17 - 10 - 2 - 2 + 19 +800e + 8 + 21 + 3 + 14 + 15 + 16 + 7 TABLE 25— Continued County Tot Dys From Cty of Origin Dys in Cty of Origin0 Out Flow Days Dys in Cty of Destin^ In Flow Days Net Gain or Loss Percent Gained Or Lost 25 9 21 13 16 17 19 38 55 55 20 36 37 37 9 33 37 10 10 14 31 10 28 29 33 + 22 7 + 9 + 2 - 41 + 12 + 9 + 25 + 49 + 53 + 12 + 31 + 13 + 29 - 6 + 25 + 37 + 1 6 + 13 + 28 8 + 21 + 28 + 27 + 63 - 8 + 13 + 3 - 26 + 71 + 14 + 24 + 91 +279 + 24 +111 + 10 + 38 - 8 + 78 + 74e + 2 - 12* + 50 +117 - 13 + 78 +215e + 61 Region II Alcona Alpena Antrim Arenac Bay Benzie Charlevoix Cheboygan Clare Crawford Emmet Gladwin Grand Trav. Iosco Isabella Kalkaska Lake Leelanau Manistee Mason Mecosta Midland Missaukee Montmorency Newaygo 35 93 69 61 157 17 63 105 54 19 50 28 125 76 75 32 .5 60 51 26 24 62 27 13 44 33 77 57 50 100 12 53 92 48 17 42 23 101 68 60 24 - 51 35 25 21 44 20 12 38 2 16 12 11 57 5 10 13 6 2 8 5 24 8 15 8 0 9 16 1 3 18 7 1 6 57 86 78 63 116 29 72 130 103 72 62 59 138 105 69 57 37 61 45 39 52 54 48 41 71 - — - TABLE 25— Continued County Tot Dys From Cty of Origin Dys in Cty of Origin0 Out Flow Days Oceana Ogemaw Osceola Oscoda Otsego Presque Isle Roscommon Wexford 62 28 48 12 80 37 77 51 59 26 39 12 61 33 67 43 3 2 9 0 19 4 10 8 1,761 1,443 318 Total In Flow Days Net Gain or Loss Percent Gained Or Lost 75 96 67 37 97 46 222 78 16 70 28 25 36 13 155 35 + 13 + 68 + 19 + 25 + 17 + 9 +145 + 27 + 21 +243 + 40 +208e + 21 + 24 +188 + 53 2,462 1,019 +701 + 40 13 15 1 3 3 5 3 3 13 7 3 9 0 2 8 2 9 + 3 38 26 92 7 1 + 4 + + + Dys in Cty of Destind Region III Allegan Barry Berrien Branch Calhoun Cass Clinton Eaton Genesee Gratiot Hillsdale Huron 70 61 56 19 32 11 73 49 273 38 30 47 57 47 47 14 20 9 32 20 168 24 26 42 13 14 9 5 12 2 41 29 105 14 4 5 70 62 48 17 23 14 35 23 181 31 29 51 + - — - - - - - 3 14 11® 28 27e 52 53 34 18 3 9 r TABLE 25— Continued County Dys in Cty of Origin0 Out Flow Days Dys in Cty of Destin^ In Flow Days Net Gain or Loss 124 20 67 57 214 71 20 38 113 8 81 128 279 102 172 70 36 78 25 101 29 47 141 53 15 45 45 143 58 15 26 53 6 58 97 169 73 95 57 18 62 23 70 21 27 40 71 5 22 12 71 13 5 12 60 2 23 31 110 29 77 13 18 16 2 31 8 20 101 58 19 49 46 162 70 23 37 62 9 71 104 180 81 99 66 20 71 23 79 22 35 43 5 4 4 1 19 12 8 11 9 3 13 7 11 8 4 9 2 9 0 9 1 9 3 Total 2,780 1,775 1,005 2,012 237 -766 TOTAL 5,519 4,076 1,443 5,519 1,443 0 + + - 66 1 18 11 52 1 3 1 51 1 10 24 99 21 73 4 16 7 2 22 7 12 98 Percent Gained Or Lost + + - 53 5 27 19 24 1 15j 3 45 13 12 18 36 21 42 6 44 9 8 22 24 26 70 - 28 113 Ingham Ionia Jackson Kalamazoo Kent Lapeer Lenawee Livingston Macomb Monroe Montcalm Muskegon Oakland Ottawa Saginaw Sanilac Shiawassee St. Clair St. Joseph Tuscola Van Buren Washtenaw Wayne Tot Dys From Cty of Origin TABLE 25— Continued The multipliers used to expand the sample data were based on the number of "usable responses" received for each of the eleven sub-regions described in Sample Design, above. For a description of the weights see Appendix G. °Total number of days generated by snowmobile owners from the county of their origin (residence). Number of days that owners spent snowmobiling in their county of origin (residence). Number of days spent snowmobiling in a county, by snowmobile owners throughout the state. The number of usable responses was less than 10. ^The number of usable responses was less than 20. 115 and nonresidents constituted the number of days at the county of destination. The contribution made by non­ residents was considered to be the number of "in-flow" days. The net gain or loss of snowmobile activity for a county amounted to the difference between the number of in-flow and out-flow days (see Figure 3). centage gained or lost was also calculated The per­ (see Figure 4). These last two columns provided a summary of the patterns of flow in and out of the counties and of the regions. However, caution was needed in interpreting the resulting estimates for those counties which had small samples, particularly those of less than ten respondents. It was interesting to examine the expanded data for the three regions. A net gain accruing to a region represented the amount of snowmobile activity done in that region by residents from one or both of the other two regions. The converse was true for a net loss. In addition to the calculations presented in Table 25, it was determined that Region I gained 1 per­ cent, Region II gained 13 percent, and Region III lost 14 percent of the snowmobiling they generated. Even though it was estimated that the residents of Region III only generated forty-two days per owner, compared to fifty-six and fifty-four days for Regions I and II, respectively, they did account for 50 percent of the total amount of use in the state. 116 +4 _ JmomioI? 1 -7 ' REGION I . t t f S T T L . l | +9 , 1t®_?»M«o [ »*T*I« 1 ^ 1 , J +9 I +16 ; + 2 7 J -7 ' 'u5utt|c ',i^,7oieM« M N Z IC m + 12 o .TRAV. 1+13 I » ; +26 i+52 , j +25 ; +23 iumftnfMi ^ 7 ^ ^ } *o7aoi>7io*ta)i**nsico REGION II /V M M ' O K « O L * I c L t a i r» L * « n « T 'L U | » I I I ,+12 ;+37 ;+19 ;+49 ; +31r , 'O C O IU y « » « * » 0 ' n i e s i r i I +141 I '+28 ! -6 ! -8 f REGION III Fig. 3. Net gain or loss, in thousands of snowmobile days. 117 , +19 jitciwioSJ \ l i “W - 1 0 «■ REGION I L— f""» No»rwM.. | +13 ; +21 | +215 r - 8 o«oor A N T ffl« t f M Z IIjfln o VBAW/ I + 71 f +10 J+78 I +279 ; +208^+63 I ioicoiT1 J-12 REGION II i +27 J+78 +a ; +40 ; +91 ; + m * # « « * ■ » 'mooim ' ( +21 ;+61 !+H7j -6 V n r ^ B S n -12 m i ic ic a 1+188 | +243 * -8 LtM~ y«5*ar«:5.I +50; i ;»*«»« * [tf / M -2^vX ! -1 8 ! -42 r - i - Hjimih i S L IK T M ) 1 REGION III Fig. 4. mobile days. Percentage of gain or loss in snow­ Insufficient data to make calculations. 118 A number of factors appeared to be operative in determining the above use patterns, such as: (1) the amount of snowfall received, or perhaps of greater sig­ nificance, the number of days for which there was some minimum number of inches, (2) the system of highways in the state, particularly in the Lower Peninsula, distribution of the snowmobile population (3) the (by county of residence), (4) the availability and suitability of the land resources, and (5) the wide range of social factors associated with snowmobile outings. The above combination of factors appeared to favor Region II as a desirable snowmobiling area for many Michiganders. Most of its counties received over sixty inches of snow in 1969-70 and many had approxi­ mately 100 inches (see Appendix E ) . There were five major highway systems leading to this region which orig­ inated in the populated areas of Region III Figure 2). supply (see Federal and state owned land was in large (see Appendix H ) . There were also numerous resort areas that had been developed for winter skiing and for summer vacationing. The attractiveness of Region II is also illus­ trated by the fact that 4 5 percent of the snowmobiling activity took place there even though only 25 percent of the registered snowmobilers resided there. As pointed out above, much of this additional activity originated 119 in the more urban region where 57 percent of the snowmobilers lived but where only 36 percent of the snow­ mobiling was done. Nonresponse The response rate to the mailed questionnaire was considered high. However, (70%) it was deemed important that some steps be taken to estimate whether or not the 30 percent who did not respond were significantly dif­ ferent from those that did reply. Therefore, telephone interviews were conducted in the counties of Ingham and Kent. A random selection of both respondents and non­ respondents was made. Forty-eight nonrespondent and thirty-nine respondent interviews were conducted in Ingham County; in Kent County forty-three nonrespondent and thirty-five respondent interviews were carried out. The responses from these two sets of interviews were compared on three variables: and amount of snowmobile use. education, income, Igo found that there were no significant differences between the nonrespondents and the respondents based on the information obtained for these three questions.^ Alison Igo, "An Analysis of the Validity of Mail Surveys for Use in Recreation Research" (Master's Thesis, Michigan State University, 1971). 120 Based on the findings of Ingo's analysis it appeared reasonable to assume that in this survey the nonrespondents were not significantly different from the respondents. Phase II In Phase II, the AID computer program was used to examine the relationship between the dependent variable, days spent snowmobiling, and selected independent vari­ ables. A description of these sixteen predictor variables and their associated classes is presented in Table 26. The goals were to: (1) determine the amount of variation in the dependent variable that was accounted for by these predictors and (2) examine the results tree to gain greater insight into the more prominent relationships that existed between the variables. The results of the algorithm were presented in the form of a tree, one for each region. were numbered, Created groups in pairs, according to the order in which the splits occurred. The letter in the lower right-hand corner of the box of the final groups indicated why they were not split further. Groups labelled E did not contain the fifty observations required to be eligible. Attempts were made to split groups labelled R; however, no pre­ dictor was able to meet the reducibility criteria. In 121 TABLE 26 PREDICTOR VARIABLES AND CLASSES USED IN THE AID ANALYSIS Variable No. 2 Variable Classes Ordering Constraints Weights (see Appendix J) 3 4 5 6 7 8 9 10 Age of Head of Household 1. Under 25 yrs. 4. 45-54 yrs. 2. 25-34 yrs. 5. 55-64 yrs. 3. 35-44 yrs. 6. 65 & over Occupation of Head . of 1. Professional 6. 2. Self-employed 7. 3. Clerical/sales 8. 4. Skilled 9. 5. Semi-skilled Household Service Farm Operators Retired Other Household Education of Head of : 1. 0-8 yrs. 4. 13-14 yrs. 2. 9-11 yrs. 5. 15-16 yrs. 3. 12 yrs. 6. 17 yrs & over Income of Household 3. Under $6000 6. 4. $6,000-$9,999 7. 5. $10,000-14,999 8. 9. $15,000-19,999 $20,000-24,999 $25,000-29,999 $30,000 & over Order Maintained Free Order Maintained Order Maintained Age Range of Children 0. Zero 2. Under 12 yrs. 1. Under 6 yrs. 3. Under 19 yrs. Order Maintained Horsepower 1. Under 16 hp. 2. 16-20 hp 3. 21-25 hp 4. 26 hp & over Order Maintained 3. Three years 4. Four years & over Order Maintained Age of Snowmobile 1. One year 2. Two years Owner of Snowmobile 1. Head 4. Daughter 2. Wife 5. Brother 3. Son 8 . Other Free 122 TABLE 26— Continued Variable NO. 11 12 13 14 15 16 17 18 Variable Classes Number of Snowmobiles Owned by Household 0. Zero 3. Three 1. One 4. Four & over 2. Two First Years of Snowmobiling 3. Before 1966 6. 1968 4. 1966 7. 1969 5. 1967 8. 1970 Snowfall 1. 30-44 2. 45-59 3. 60-74 4. 75-89 in County in. in. in. in. of 5. 6. 7. Residence 90-104 in. 105-119 in. 120 in.& over Ordering Constraints Order Maintained Order Maintained Order Maintained Present Regulations Should Be: 1. More strict 6. Less strict 4. Unchanged Free Enforcement of Regulation Should Be: 1. More strict 6. Less strict 4. Unchanged Free Regulations Near Ice-fishing 1. Yes 6. No. Free Club Membership 1. Yes Free 6. No Preferred Snowmobile Group 1. Alone 5. Organized 2. Wife group 3. Children 6. Family 4. Friends 7. Relations Free 123 other w o r d s , no one variable explained enough of the variation for the BSS /TSSm to meet the criterion set p T for each region. Region I It is important to note, according to Sonquist, the variables included in the tree and the order in which they appeared. Those variables with high predictive power tended to be used early in the analysis; those which affected small sub-groups appeared in later splits.1 Another property of the tree was that of nonsymmetry, in terms of the extent and manner in which variables were used in the various branches, which indicated the pre­ sence of interaction effects. If a variable operated in a different manner in different branches or if it only had utility, or potential utility, in reducing predictive error in one branch then it was a clear indication that such a variable interacted with those used in preceding splits.^ The results tree obtained for Region I using the predictor variables and the program restrictions outlined above was presented in Figure 5. This tree was initially split into two main branches, on the number of snowmobiles 1Sonquist, Multivariate Model Building, p. 96. 2 Ibid.; Sonquist and Morgan, Detecting Inter­ action Effects, pp. 111-12. Head: grade 8 or less Y = 8? (1 8 ) N = 21 Head: 34 years o r under_ E (1*) Head: 54 years o r under Two o r more snow­ mobiles ner household 0?) Y = 55 Head: 35 years to 54 years Y = 62 (1 5 ) N = 107 24% Head: grade 9 o r more (11 ) 3% Head: 55 years or over Y = 37 (1 3 ) N = 17 Region I snownobile owners y = 5fi (1 ) N - jj° E E R E - - Group d id not have 50 o bservations R - - Group f a i l e d to s a t is f y r e d u c i b i li t y c r i t e r i o n ( .0 1 0 ) N = 523 23% 21 hD o r over Head: n ot s e l f Owned by w ife , daughter_or o th e r 37% One Snowmobile oer household 20 hp o r under Owned by head o r son Head: self-em p loyed Fig. 5. Region I . Reduction in E rro r BSS /TSS = .142 AID "results tree": an analysis of the number of snowmobiling days in owned by the household. This produced a high-use group that owned two or more snowmobiles and a low-use group having but one machine (this group included two respondents who had disposed of their snowmobile). After this initial split, this predictor lost its power to reduce the variation in the dependent variable. The occupation variable was found to be the most powerful predictor for group 2, the low-use group, to split on; no other predictor came close, even though six predictors satisfied the criterion (see Table 27).'*' It was used later in an attempt to split group 7, but failed since it did not satisfy the reducibility criterion. Occupation also demonstrated potential strength in the top branch of the tree. Horsepower retained the strength it had demon­ strated in group 2 and was used to split group 4. How­ ever, after this split it gave no further evidence of po wer. The use of snowmobile ownership to partition group 6 was of special interest as its use was very likely spurious. The group was badly skewed on this variable, 88 percent of the group being heads of households. ■*"This table did not indicate the proportion that the between sum of squares was of the total sum of squares (BSS /TSST ) . This proportion was calculated for each of the regions and placed in Appendix K. Statistics for the splits which were attempted but failed were not included in the table. TABLE 27 REGION I: PROPORTION OF VARIATION IN EACH GROUP EXPLAINED BY EACH PREDICTOR VARIABLE3 Parent Group : Number*3 Predictor Age Occupation Education Income Age Range Horsepower Age of SnowM. Owner No. of SnowM. Years of SnowM. Snowfall Regulations Enforcement Ice-fishing Club Member Preferred Grp. N Mean TSSi/TSST 1 2 3 4 6 .023 .026c .015 .008 .023 .012 .003 .013 .034d .002 .005 .014 .020 .000 .002 .012 .018 .043d .014 .010 .015 .019 .011 .009 .001 .007 .003 .016 .020c .002 .001 .016 .047° .044 .052d .043 .021 .011 .022 .013 .000 .009 .022 .006 .010 .007 .000 .030 .011 .008 .018c .013 .017 .022d .015 .009 .001 .005 .009 .009 .017 .001 .007 .017 .011 .012 .020 .032 .022 .005 .033° .036d .002 .008 .018 .008 .012 .000 .012 .022 .092d .042c .008 .024 .029 .014 .036 .024 .013 .012 .014 .004 .008 .003 .001 .035 .060d .044c .030 .021 .005 .004 .033 .038 .006 .006 .029 .006 .004 .008 .000 .010 523 356 167 313 199 146 129 56 52 64 55 51 62 65 .674 .293 .598 .380 .247 .209 1.00 aProportion of variation in group i_ explained by each predictor 11 12 (BSS/TSS)^. The group numbers correspond to those in the lower left-hand corner of the boxes in Figure 5. cPredictor which contained the second largest (BSS/TSS)^. ^Predictor on which the partition was made. 127 Therefore, it was primarily the shape of this predictor that was responsible for the split rather than the influence of its components in reducing the predictive error. In the top branch, group 3 was split next in the iteration on education. This isolated a small group of high users that had less than grade 9 education. This predictor was used in an attempt to split group 15 and was the second best predictor in group 4 in the bottom branch. Age was the last variable to be used, producing splits in groups 11 and 12 in succession. Only six of the sixteen predictor variables were used in this analysis; however, several other predictors came close to being used. Income and age of snowmobile each demonstrated potential in both branches; both came close to partitioning group 6. for example, Attitude toward enforcement of snowmobile regulations, age range of chil­ dren, and preferred snowmobile groups all shewed potential power in the bottom branch. The analysis produced a reduction in the error variance (BSS/TSS) of 14.2 percent. An examination of ^"In discussions with J. Paul Johnston, Associate Professor of Political Science, University of Alberta, June 1974, he pointed out that when a predictor was involved in successive splits on one branch, usually it was responding to the nonlinearity present in the group on that predictor. 128 the final groups (see Table 28) showed that three of them still contained a large number of observations (over 100). These remained as "unexplained" groups as no predictor was capable of reducing the unexplained variance the required amount.^ Region II A more complex pattern of interaction effects was observed in the results tree of Region II (see Figure 6). The initial parent group was split, as in Region I, on number of snowmobiles. This partition used up this pre­ dictor's utility in both branches (see Table 29). Age range and snowfall, which were the second and third best predictors on the initial group, were used alternately in the next four splits. This process exhausted the utility of age range; however, snowfall was able to maintain a moderate relationship with the depen­ dent variable in the bottom branch. Snowfall also acted, to a degree, as a proxy for geographic location, for example, those respondents from counties receiving less than seventy-five inches of snow (group 10) resided in one of the ten counties in the southeast corner of the region (see Appendix E ) . ^"Sonquist and Morgan, Detection of Interaction Effects, pp. 50-51. To reduce the error in these groups would have required: (1) lowering the reducibility cri­ terion, or (2) introducing other predictors into the analysis, or (3) some combination of these two alterna­ tives . 129 TABLE 28 REGION I: CHARACTERISTICS OF THE FINAL GROUPS CREATED IN THE AID ANALYSIS3 Characteristics of Groups Group N Mean SD Owned Two or More Snowmobiles and: 10 Head had grade 8 or less 21 82 27 14 Head had grade 8 or less, and was under 35 years old 22 80 29 Head had grade 9 or more, and was 35 to 54 years old 107 62 22 Head had grade 9 or m o r e , and was 55 years or older 17 37 22 Head was not self-employed; snowmobile had less than 21 hp, and was owned by wife or daughter 13 72 30 Head was not self-employed, and snowmobile had 21 hp or more 114 60 29 Head was not self-employed; snowmobile had under 21 hp, and was owned by head or son 186 50 30 43 35 23 523 56 31 15 13 Owned One Snowmobile and: 8 7 9 5 Head was self-employed Total Sample £ The final groups were arranged under the two main branches of the tree and in descending order of their means. Snowf al l 75" 4 over i n r e s i d e n t county Y = 80 ( 1 1) N - 95 R Snowf al l 74" 4 under i n r e s i d e n t county C h i l d r e n 1? to 18 y e a r s i n household (9) Y = 74 ( 1 0) Y = 6-1 N = 61 E - - Group di d not have 50 o b s e r v a t i o n s R R - - Group f a i l e d to satisfy redu cib ility c r i t e r i o n (.009) Owned by: w i f e or son Y = 90 Two or more snow­ mobi l es per household Y = F6 ( 1 6) C h i l d r e n under 12 ye a r s or no c h i l d r e n (8) Reoion I I owners (i) Y = 57 K 1a- N = 8 E Owned by head or other Y = 55 ( 1 7) N = 113 Combined income 5 1 0 , 0 0 0 or over Y = 69 ill}------------ t L = 23------ R snowmobile R Y = 55 N = 825 Snowf al l over °n" i n r e s i d e n t county Combined income Less than 81 0,0 0 0 (7) (12) Y = 60 One snowmobile oer household (5) Y = 53 S no wf a l 1 under 90" i n r e s i d e n t county f 61 No c h i l d r e n i n household (4) Y = as N = 231 Cl ub member Re g u l a t i o n s near ice-fishing Y = 47 (19) N = 66 Y = 7° ( 2 0) R N = 9 E Reduct i on i n e r r o r BSS /TSS = .203 No r e g u l a t i o n s near i c e - f i s h i ng Non cl u b member ( 1 4) (21 ) Y = 45 Y = so N = 53 R Y = 40 Re g u l a t i o n s near i c e - f i s h i ng (15) Profession, s k i l l , s k i l l O' f ar mer Y = 43 ( 2 2) tl = 46 Y = 33 N 5' AID "results tree": semi ­ E Self- employ, se rv i ce , s a l e s , r e t i r or o t h e r (23) Fig. 6. Region II. < Y = 53 Y = 24 N = 43 E an analysis of the number of snowmobiling days in 130 C h i l d r e n 1 t o 18 y e a r s i n household No r e g u l a t i o n s near ice-fishing Y = 67 (18) N = 26 E R TABLE 29 REGION II: PROPORTION OF VARIATION IN EACH GROUP EXPLAINED BY EACH PREDICTOR3, Parent Group Number*3 Predictor N Mean TSSi/TSST 2 3 4 5 7 8 9 .027 .018 .003 .005 .049c .002 .002 .010 .064d .006 .029 .022 .014 .021 .019 .008 .030 .024 .001 .004 .042 .003 .013 .007 .002 .012 .030 .009 .010 .032c .026 .009 .036 .019 .022 .026 ,070d .001 .003 .010 .003 .001 .033 .039° .008 .003 .001 .010 .036 .044 .002 .018 .000 .038 .026 .036 .002 .014 .021 .014 .045 .089d .083c .041 .014 .011 .002 .008 .006 .002 .019c .010 .002 •013 . .0 39d .014 .001 .019 .016 .011 .016 .030 .041c .070d .029 .006 .038 .015 .000 .001 .007 .019 .006 .040 .011 .015 .060c .031 .016 .034 .006 .010 .008 .068 .005 .006 .019 .032 .009 .021 .020 .044 .018 .024 .025 .021 .000 .004 .011 .006 .003 .001 .062d .062° .006 .010 .004 .010 825 547 278 151 396 165 121 157 92 62 89 55 49 66 40 53 60 57 74 53 50 33 .596 .340 .154 .417 .172 .136 .180 .086 .076 .065 1.00 aProportion of variation in group i^ explained by each predictor ^The group numbers correspond to those used in Figure 6. cPredictor which contained the second largest ^Predictor on which the partition was made. (BSS/TSS)^. (BSS/TSS)^. 12 .033 .056 .013 .004 .019 .032 .061 .020 .000 .017 .024 .022 .008 .113d .000 .087c 14 .015 .121c .045 .088 .000 .002 .064 .094 .000 .028 .109 .082 .007 .000 .152d .077 15 .143° .170d .016 .035 .000 .132 .048 .006 .002 .023 .031 .045 .114 .000 .025 .063 131 Age Occupation Education Income Age Range Horsepower Age of SnowM. Owner No. of SnowM. Years of SnowM. Snowfall Regulations Enforcement Ice-fishing Club Member Preferred Grp. 1 132 Group 7 was partitioned on income; $10,000 acting as the cutting point. This predictor also exhibited potential utility on group 3 of the top branch. Attitude toward regulations near ice-fishing was used next to split group 4. Respondents who were opposed to the establishment of regulations limiting snowmobile activity near ice-fishing did more snowmobiling. This predictor was also used later to split group 12, but was never useful in the top branch. The ownership variable came forth again, as it did in Region I, to produce a spurious split (for the same reasons as stated above) on group 8. Age, which was not used in Region II, was a close second-best in this situation and was also used in an attempt to split group 17. Age also demonstrated potential power in splitting the groups on which age range was used. Membership in a snowmobile club only exhibited power in the bottom branch, group 14. finally being used to split Being a club member was associated with high snowmobile use. Even though occupation was the last predictor to be used, splitting group 15, it exhibited potential utility throughout the tree. If the reducibility cri­ terion had been set at a lower level then this variable would have been used to split groups 13 and 16. 133 The analysis reduced the amount of variation by 20.3 percent. Of the twelve final groups only two were considered to be very large (see Table 30). Region III In Region III, as in the other regions, the initial parent group was split on number of snowmobiles (see Figure 7).^" However, in this region, this predictor did exhibit some power in the later stages of the tree (see Table 31). The predictors, age and occupation, demonstrated greater utility in this region, as both were used to split groups on three occasions throughout the analysis and showed strength in two other instances. cation of a high degree of interaction, As an indi­ their mode of operation was different in each instance. Snowfall, as in Region II, was a powerful pre­ dictor, being used twice to split groups, showing strength on two other occasions and was selected twice to attempt splits on final groups (groups 16 and 22). However, the interaction picture was not clear as on the two occasions in which it was used the split occurred at the same point, but no information was available from the analysis as to its manner of operation in the other ^"Group 2, one snowmobile households, also included eight households in which the owners had disposed of their snowmobile. 134 TABLE 30 REGION II: CHARACTERISTICS OF THE FINAL GROUPS CREATED IN THE AID ANALYSIS3 Group Characteristics of Groups N Mean SD 8 90 18 Household had children 12 to 18 years old, and county received 75 inches or more of snow 96 80 29 Household had children 12 to 18 years old, and county received less than 75 inches of snow - 61 64 32 113 55 16 Owned Two or More Snowmobiles a n d : 16 11 10 17 Household had children under 12 years old or no children, and first reported snowmobile owned by wife or son Household had children under 12 years old or no children Owned One Snowmobile and: 20 13 18 6 Household had no children; respondent was opposed to regu­ lations near ice-fishing, and was a member of a snowmobile club 78 25 Household had children 1 to 18 years old; county received 90 inches or more of snow, and household earned $10,000 or more 73 69 29 Household had children 1 to 18 years old; county received 90 inches or more of snow; household earned less than $10,000, and respondent was opposed to regulations near icefishing 26 67 28 Household had children 1 to 18 years old; and county received under 90 inches of snow 231 48 29 135 TABLE 30— Continued Group Characteristics of Groups N Mean SD Owned One Snowmobile and: 19 21 22 23 Household had children 1 to 18 years old; county received 90 inches or more of snow; household earned under $10,000, and respondent favored regulations near ice-fishing 66 47 25 Household had no children; respondent opposed regulation near ice-fishing, and did not belong to a snowmobile club 53 45 30 Household had no children; respondent favored regulations near ice-fishing, and head was a pro­ fessional, skilled, semi-skilled or farmer 46 43 25 Household had no children; respondent favored regulations near ice-fishing, and head was selfemployed, clerical/sales or retired 43 24 17 825 55 31 Total Sample aThe final groups were arranged under the two main branches of the tree and in descending order of their means. S n o w fall in 6 0" resid en t o r over S elf-em ploy , s k i l l , s a^ es semi or «erv. 67 Y - 56 J2 2 1 ----------- 3 _ i . 9 c ounty Y = 60 pr c.f e ss ' o n , s k i l l , fa rm er, - e t i r e d nr o ther -v _ cr Two o r ^ o r e snow­ m obiles h o u s e h o ld oe- ^ Head: unoer S n o w fa ll in 6 9" re s id e n t * C h i l d r e n 1 t o 18 years in h o u s e h o ld Y = 5° (2 7 ) N = 41 ____ ?.1 vears o*' No c h i l d r e n i n h o u s e h o ld y - 37 (26) ri - 15 urder M _ » county S n o w fa ll in 6 0" re sid e n t (Hi He ad : over ?5 v e a r s (1 3 ) ? e n ’o n ITT ?nov>r,oh'? le owner s 7 = E Y = <33 Y s d6 fl) E or over county V = 53 Head: 3 * under years (20 ) l| = 57 or Y = 61 or v=-S U = 3^1 A3 S n o w f a l l 5 5" ’ u n d e r in r e s i d e n t countv He ad : 35 V =.'.2 ( 1n ) 'I = 251 to ff years Y = 43 5 (2 1 ) P = 47 N = 1*31 S k ill, farm er s e1" ’ - S k ' 1 1 , or o th er u‘e a d : q r a d e 8 o r ( 1°) Head: 4 4 y e a r s under M l less Y = SO i* = 5 c or 6 rade ~ or Y = 43 over Y = 37 r rcfess en n 'o y, se^vce on, s e l f sa ; es o r N - ^ E — o Combined Inco me 520 ,000 o r over Y = 37 !7) One s n o v ^ o b M e o e r h o u s e h o ld (1 9 ) Farm er, re tire d or (M ) He ad : over *5 years Y - Combined les s o th er Y = <*3 ULi or «°n than h ro fo s s ’ on, enr 1o / , s e lf - s'*' 1 1. se'” ’* s k ill, ser/, or Gr oup f a i l e d t o s a tis fy r e d u c ib ility c rite rio n (.0 0 6 ) t income S2 r,, nrJr; Y = 38 M = -53 Reduction in Error BSS /TSS = .151 " i r s t ourchase b e f o r e 1 96 8 33 n o t h a ve 50 o b s e r v a t i o n s — Y = 73 Z51 Group d i d Y = 5^ - U Z J ---------- !1 = 11 sale - i r s t purchase 1*563 o r l a t e r Y - 30 - Fig. 7. Region III. AID "results tree": ?Qf, o an analysis of the number of snowmobiling days in E TABLE 31 REGION III: PROPORTION OF VARIATION IN EACH GROUP EXPLAINED BY EACH PREDICTOR3 Parent Group Number** Predictor N Mean TSSi /1 S S T 3 4 5 6 7 8 9 .026° .009 .004 .003 .014 .004 .001 .007 .034“ .003 .022 .005 .001 .002 .008 .021 .035d .022 .004 .005 .011 .009 .004 .002 .001 .005 .015 .003 .004 .006 .013 .032c .013 .015 .012 .012 .016° .005 .006 .012 .015 .003 •063d .006 .002 .000 .000 .005 .013 .023d .003 .009 .003 .008 .002 .003 .003 .004 .013 .004 .001 .002 .010 .023c .008 .040d .005 .008 .006 .009 .004 .003 .000 .033 .027 .004 .018 .021 .017 .035“ .017 .005 .001 .005 .001 .013 .003 .005 .004 .010 •036d .003 .010 .000 .006 .023“ .020 .003 .054d .032“ .001 .007 .011 .014 .003 .001 .008 .016 .021 .006 .012 .030 .029d .015“ .007 .003 .004 .004 .013 .013 .007 .003 .001 .006 .004 .001 .004 .012 .033 .081 .007 .066 .068“ .022 .002 .059 .037 .008 .012 .019 .008 .001 .019 .022 .002 .010 .005 .002 .003 .021“ .006 .002 .002 .017 .000 .013 .011 .001 .002 .042d .102d .011 .019 .012 .023 .044c .015 .008 .022 .003 .004 .032 .025 .015 .026 .044 .009 .013 .021 .20 2d .035 .027 .012 .024 .000 .021 .015 .183“ .014 .023 .004 .093 .011 .009 .006 .005 .001 .007 .006 .010 .000 .049d .031 .017 .018 .022 .022 .039c .114“ .055 .033 .081 .178d .056 .022 .068 .105 .022 .014 .024 .024 .004 .037 .029 1431 964 467 594 370 365 229 355 112 261 104 63 307 56 31 39 50 43 33 46 38 46 60 42 53 43 31 54 .659 .307 .434 .202 .299 .125 .213 .075 .187 .101 .052 .142 .034 1.00 P r o p o r t i o n of variation in group i explained by each predictor (BSS/TSS)^ bThe group numbers correspond to those used in Figure 7. £ Predictor which contained the second largest (BSS/TSS)^. P r e d i c t o r on which the partition was made. 11 14 23 2 137 Age Occupation Education Income Age Range Horsepower Age of SnowM. Owner No. of SnowM. Years of SnowM. Snowfall Regulations Enforcement Ice-fishing Club Member Preferred Grp. 10 15 1 138 demonstrations of power. It was observed that 73 percent of the total number of respondents resided in counties receiving less than sixty inches of snowfall. Three other variables, age range, education, and income, were used once and were second-best on one other occasion. The number of years a respondent had been snowmobiling was introduced for the first time; it was used once and was third-best on another occasion. The preferred group predictor was of special interest. It produced a split on group 10 but was deleted from the tree due to its spurious nature. Not only was it a free predictor but it also contained a large number of classes, several of which overlapped. The analysis reduced the predictive error by 15.2 percent. An examination of the final groups (see Table 32) pointed out that four contained a large number of observations. Summary In each analysis the selected predictor variables explained a relatively small proportion of the variation (20% or less) . The presence, in each analysis, of a number of large final groups which still contained a large amount of variation (unexplainable groups) indi­ cated that either additional variables were needed or 139 TABLE 32 REGION III: CHARACTERISTICS OF THE FINAL GROUPS CREATED IN THE AID ANALYSIS61 Group Characteristics of Groups N Mean SD 4 83 14 County received over 60 inches of snow, and head was self-employed, semi-skilled or clerical/sales 56 67 24 County received over 60 inches of snow; head was professional, skilled, farmer or other, and household had children 1 to 18 years old 41 59 22 351 45 25 15 37 12 5 80 31 Head was 45 years or over, a farmer, retired or other, and household earned $20,000 or more 10 73 33 Head was under 35 years, a skilled, semi-skilled, farmer or other, and county received 60 inches or more of snow 57 61 28 Owned Two or More Snowmobiles a n d : 12 22 27 13 26 County received under 60 inches of snow, and respondent was under 25 years old County received under 60 inches of snow, and head was 25 years old or over County received over 60 inches of snow; head was professional, skilled, farmer or other, and household had no children Owned One Snowmobile a n d : 18 25 20 Head was under 4 5 years old; a professional, clerical/sales or self-employed, and had grade 8 or less 140 TABLE 32— Continued Group Characteristics of Groups N Mean SD Owned One Snowmobile and: 17 21 10 24 19 16 Head was 45 years or over; a pro­ fessional, skilled, semi-skilled, clerical/sales or self-employed, and started snowmobiling before 1968 11 54 26 Head was 35 to 44 years old; a skilled, semi-skilled, farmer or other, and the county received 60 inches or more of snow 47 43 23 261 42 26 53 38 22 Head was under 4 5 years; a pro­ fessional, clerical/sales or selfemployed, and had grade 9 or more 224 37 24 Head was 4 5 years or over; a pro­ fessional, skilled, semi-skilled, clerical/sales or self-employed, and started snowmobiling in 196 8 or later 296 30 21 1431 43 26 Head was under 4 5 years; a skilled, semi-skilled, farmer or other, and county received under 60 inches of snow Head was 4 5 years or over; a farmer, retired or other, and household earned under $20,000 Total Sample The final groups were arranged under the two main branches of the tree and in descending order of their means. 141 the present ones should be restructured in order to reduce the predictive error.^ The nonsymmetry exhibited by the three trees implied that interaction effects were present, that is, effects of a combination of factors. In other words, when a predictor was used in one trunk of the tree and did not exhibit actual or potential utility in the other branch, then this indicated that there was an interaction effect present between that predictor and those used in the preceding splits. 2 Also, as pointed out above, when a variable was partitioned in a different manner in one branch compared to another, it was usually responding to the nonlinearity present in that group on that predictor. Only in the case of the variable "snowfall" in the analy­ sis of Region III was it observed that a predictor behaved in the same manner in both trunks of a tree. Where only one predictor showed up as being important in producing a split, then it was unlikely that its predictive power was a result of sampling variability. However, when several predictors displayed similar strength in their ability to partition a group then the chances were greater that the choice of one ^"Sonquist and Morgan, Detection of Interaction E f f e c t s , pp. 110-13. ^Ibid. 142 over another was due to sampling variability.^ For example, in Region II "age of household head" was second best on two occasions but was never used to split a gr oup. Interaction effects between predictor variables is a complex phenomenon and its assessment requires con­ siderable information about the variables concerned. In order to gain greater insight into these effects, it would be useful to obtain additional information, such as: (1) the distribution of the variables in the parent groups, particularly since variables may become skewed in the partitioning process and (2) the frequency dis­ tribution of the residuals to provide information about the extent that one variable substitutes for another. ^Ibid., pp. 124-25. CHAPTER IV CONCLUSIONS AND RECOMMENDATIONS Context of the Study At the time this study was conducted, snowmobiling was still a relatively new recreation activity. It was not until the late 1960's that it began to gain wide­ spread popularity in Michigan. Initially, it gathered adherents in the Upper Peninsula and then spread south into the more urban regions. The state of Michigan has several characteristics that present special problems for recreation land manage­ ment agencies. In particular, the concentration of two- thirds of the population in the southeast corner of the state while the great majority of public recreation lands are in the north and west portions of the state places great pressure on the state parks and recreation areas near the densely populated urban areas. The task of providing recreation resources for snowmobiling is further complicated by the shortage of snowfall in the southeast portion. This study has three major thrusts. The first is concerned with the methods of selecting data for a study 143 144 of this magnitude about snowmobiling. Since no studies of this nature had been conducted it was considered important to examine the methods that were selected for this study. The second is to determine, on a broad base, the use patterns and user characteristics of registered snowmobile owners in the state of Michigan. The third is to consider the implications for decision­ making that the results have for planning and management agencies. When this study was conceived it was decided to conduct it on a state-wide basis. In order to improve measurement precision, the state was stratified by the three major geographical regions, as discussed earlier. The boundaries of these regions also coincided with the major divisions used by the Department of Natural Resources in planning and managing the state's resources. However, these regions are not homogeneous in some of their characteristics, such as: degree of urbanization, amount of snowfall, available public land for snowmobiling, etc. Because of this situation the results are not as useful for on-site planning and man­ agement as they are for state-wide recreation planning. Recommendation.-— In subsequent studies of snowmobilers it is recommended that the spatial units used for collecting and analyzing data should be more homogeneous in terms of the variables being investigated. 145 Data analyzed on this basis should provide greater insights into the use patterns and user characteristics which would assist in the management of the resource base. Study Methods Response r a t e .— A problem frequently encountered with self-administered questionnaires is low response. In this study it was demonstrated that a high response rate may be obtained by giving attention to details, such as: (1) obtaining sponsorship that is acceptable to the respondent, (2) carefully describing the nature of the project, using explanatory letters, (3) sending reminders to nonrespondents when the initial response rate begins to diminish, (4) mailing a second questionnaire to non­ respondents as soon as the response to reminder cards drops off, and (5) designing a questionnaire that is attractive and appears to require a short time to com­ plete . Attitude scales.— Respondents in this study were asked to respond to two types of attitude scales: (1) one that provided a range of responses including a neutral response and choice. (2) one which only offered a dichotomous It was of interest to note that where the oppor­ tunity for a neutral response was provided, over 50 per­ cent of the respondents chose such a response. This behavior would seem to imply that if respondents were 146 not given this choice then the response rate would diminish considerably. However, this was not the case in this study, as there was only a slight drop in response when respondents were only given an opportunity to answer "yes" or "no." Therefore, the best solution would appear to be that of offering respondents some range to express their opinion but to eliminate the choice of a neutral response. Measurement of time.— In this study, time spent snowmobiling was measured in "days" where any part of a day was counted as one day. This method has been used frequently in recreation surveys; however, it does present some problems in terms of precision. For example, those that went snowmobiling for many hours per day would be represented by the same number of days as those who had only spent a few hours per day on this activity. It also resulted in double counting in some instances, such as: reporting use of three classes of land used in one day would be reported as three days. Some improvement in measurement of time would be desirable. Recommendation.— In subsequent studies of snow­ mobile use it is recommended that the use of "day" as a unit of measurement be refined. In addition to reporting the number of snowmobiling days the respondent should be requested to estimate the average number of hours per day 147 that he devotes to snowmobiling and, also, to consider the proportion of time spent on each component of the variable being studied (for example, estimation of the proportion of time spent on each kind of snowmobile activity, such as: trail riding, scrambling, etc.). The use of such techniques, if thoroughly field tested, should provide more precise estimates of the amount of time spent on various snowmobile uses. Estimation of socio-economic variables.— In this study, some difficulty was experienced in estimating the socio-economic characteristics of snowmobile owners since much of the information collected was for heads of house­ holds. Eighty-seven percent of the respondents in this study were heads of households so that the socio-economic data collected may not have been exactly the same as information for the snowmobile owners. Recommendation.— It is recommended that in future studies of snowmobile owners the socio-economic data be collected for owners rather than heads of households. This would enable the researcher to more accurately describe such respondents. Detection of interaction effects.— The Automatic Interaction Detector (AID) technique was used in the analysis phase to determine the presence of interaction 148 effects among the independent variables (the dependent variable was "number of snowmobiling days") . The socio-economic variables, in particular, "occupation" and "age" of head of household, demonstrated power as "predictors" of the amount of snowmobiling undertaken, when they were used in the AID analyses.^" The variable "number of snowmobiles owned" by each household explained the largest proportion of variation in the dependent variable compared to all other predictor variables used. However, it is suggested that much of the power exhibited by this predictor was due to its action as a proxy variable for "income." In other words, the strength of income as a predictor of snow­ mobile use may have been much stronger than the analyses indicated. The number of socio-economic variables used in each of the AID analyses varied from three, in Regions I and II, to all five in Region III. The results of the algorithm indicated that interaction effects were present in a number of groupings of these variables. However, the nature and extent of these interaction effects was not determined, therefore, their roles as predictors could not be assessed accurately. ^The power of predictor variable for each parent group is the proportion of variation (BSS^/TSS) in the dependent variable that the predictor contains for that group. (As explained earlier, the term "predictor" is not used here in terms of developing a functional equation.) 149 Recommendation.— It is recommended that in future studies the nature and extent of the interaction between predictor variables should be determined. In order to accomplish this, the correlations between the variables of interest should be established. After the "parent groups" have been produced by the AID algorithm, profiles for the predictor variables across the dependent variable should be developed for each of these groups. An examination of the shape of these distributions along with correlation information will enable the researcher to understand the nature of the interaction effects that are present. User Characteristics Age level of head of household.— The heads of snowmobile-owning households in Region I were older, having a mean age of 44.4 years compared to 42.6 years and 41.7 years in Regions II and III, respectively. These patterns between the three regions were similar to those found in the general population. Fifty-nine percent of the heads of snowmobile-owning households in the state were between 35 and 54 years of age, whereas in the general population this age group only made up 34 percent of the population. Compared to the per­ centage of those 20-24 years old in the general popu­ lation, the proportion of heads of snowmobile-owning households who were 25 years and under was much smaller. 150 As was expected, only a small percentage of the general population in the 6 5 years and over category owned snow­ mobiles . Educational level of head of household.— For the state as a whole, 70 percent of the heads of snowmobileowning households had completed high school. Those in Region III had completed more years of education than the heads from the other two regions: this pattern was similar to that found in the general population. The proportion of heads of snowmobile-owning households that went beyond high school was three times as great in the samples from Regions II and III as it was in the general population. In Region I it was four times as large. Income of household.— There were marked dif­ ferences between the combined gross income of snowmobileowning households in each of the three regions. In addition, the incomes of snowmobile-owning households were considerably higher than those found in the general population. The observed differences (from $10,900 to $15,200) may have reflected, at least in part, the dis­ parities in the cost of snowmobiling in each of the regions. In other words, since it cost more to go snowmobiling in Region III, due to increased travel costs, etc., the activity tended to be restricted to members of households having higher incomes. 151 A much smaller proportion of the general popu­ lation in Region III owned snowmobiles compared to the other regions. This may have been due to the presence of a greater variety of recreational opportunities in the urban centers as well as less public land and poorer snow conditions. It is of interest, from a management perspective, to note that snowmobilers in Region III are relatively affluent, as 40 percent of the snowmobile-owning house­ holds earned $15,000 or more. Therefore, many of them are financially able to pay more than a nominal fee to gain access to land for snowmobiling. This suggests that it would be useful to examine the respective roles of public and private agencies in providing land resources for snowmobiling. Occupation of head of household.— The observed regional differences in occupation patterns and region were of a more complex nature. A large percentage of heads of snowmobile-owning households in each region were self-employed persons and managers. In comparison, a small proportion of those persons employed in clerical, sales, service, and unskilled positions were respondent household heads. These differences appear to reflect the close relationship between these occupations and income. 152 The professional, skilled, semi-skilled, and farm management occupations did not exhibit consistent patterns from region to region. For example, in Regions I and II, a larger proportion of heads of snowmobile-owning households was skilled persons than in the general popu­ lation; however, this relationship was reversed in Region III. It appeared that there was considerable interaction between these components of the occupation variable and other factors, such as income. Size of household.— A statistically significant difference was obtained in the number of children per household, 18 years old and under, in each of the regions. From the point of view of a management agency, these dif­ ferences do not appear meaningful. Respondent households of the state had an overall average of 1.8 children. However, only 71 percent of the households reported having children 18 years old and under: the average for these households was 2.5 children. Patterns of Use Amount of snowmobile u s e .— Snowmobile owners in Regions I and II spent a significantly greater number of days snowmobiling than did respondents from Region III (an average of 56 and 54 days compared to 4 2 days, respectively). However, since there were 1.3 times as many registered snowmobiles in Region III than in 153 the other two regions combined, the Region III snowmobilers generated 50 percent of the total number of snowmobiling days for the state as a whole. Twenty-eight percent of the days originating in Region III was spent outside of this region, mainly in Region II (see Table 25). While Region III experienced an outflow of snowmobiling activity, Region II received an inflow of approximately the same magnitude. This inflow of snowmobilers should have consider­ able economic impact on the region and, more particularly, on principal destination counties such as Roscommon County. That county provided a total of 222,000 snow­ mobile days in the 1969-70 season representing a 188 per­ cent increase over the 77,000 days generated by its residents. The expanded data, which only provided rough estimates at the county level, did suggest that the snowmobile use varied greatly from county to county. The reasons for this in each case was not evident; how­ ever, such information would be useful to planners and to managers of resources used for snowmobiling. Classes of land used.— In examining the classes of land used for snowmobiling, it was found that respondents in Regions II and III spent 30 percent of their snowmobil­ ing time on their own land. It is of interest that this proportion was so high for snowmobilers in Region III, many of whom reside in cities. The conclusion would 154 appear to be that snowmobilers in this region who lived in a rural setting tend to do most of their snowmobiling on their own land; however, this needs further study. Respondents throughout the state reported a majority (54%) of their time was spent on nonpublic land ("owners land" and "private land"), with only 4 2 percent on land owned by county, state, and federal governments reported use of "local roads").'*' (including Since the activity on nonpublic land constituted such a large portion of total snowmobile use, it is important for management and plan­ ning agencies to determine if it is of a different nature than the use of public lands and, if so, what is the relationship between the two use patterns? Snowmobiling and associated activities.— Snow­ mobilers spent a greater proportion of their time (58%) trail riding compared to scrambling in open areas (31%). It would be of interest from a management point of view to know to what extent these different proportions reflected the users' preferences rather than the oppor­ tunities available to them. It was noted that respondents from Region III spent 7 percent more time scrambling than those from the other regions, but it was not determined if this was related to: (1) their greater use of 1 Respondents indicated that only 0.4 percent of their time was spent snowmobiling on privately owned land for which a fee was charged. 155 privately owned land, (2) the lighter snow cover, which is more conducive to scrambling than deep snow, or (3) a greater preference for scrambling by those from an urban region. The study showed that the recreation activity of snowmobiling involved more than just "snowmobiling." Twenty-four percent of snowmobile activity was closely associated with a number of other recreation activities.^ Not only were snowmobiles used on hunting and fishing trips, where snowmobiling may have played a secondary role, but they were also involved in "outings" in which members of the household participated in such activities as skiing, tobaggoning, and cook-outs. This implies that, in many cases, the area chosen for snowmobiling is influenced by the opportunities for household members to participate in one or more of these associated activi­ ties. Therefore, management may be able to influence the amount of snowmobile use an area will receive, by either encouraging or curtailing participation in other recreation activities. The social nature of snowmobiling was also indicated by a majority of respondents stating that they preferred to snowmobile with their family or friends; only 5 percent preferred to go snowmobiling alone. ^Based on the total reported number of days spent on "associated activities" (see Table 23) compared to the total number of days spent participating in "kinds of snowmobile activity" (see Table 22). 156 Mobility.— Snowmobilers throughout the state appeared to have a high degree of mobility transport their snowmobiles). (ability to In particular, 68 percent of Region III snowmobilers reported that they owned a snowmobile trailer. In addition, 13 percent indicated that they owned a truck which could be used to transport their snowmobiles. For snowmobilers in Region III, the importance of having mobility was further illustrated by 36 percent of their snowmobiling being done outside of their county of residence compared to only 12 percent for respondents in Region I and 18 percent in Region II. ^ Much of the travel done by snowmobilers in Region III consisted of overnight trips; 39 percent of them reported taking an average of seven trips of over 100 miles from home involving two or three nights away from their residence. Snowmobilers in Region I were more inclined to take short trips; 21 percent reported taking an average of six overnight trips of less than 100 miles. However, respondents in Region II did not appear to travel extensively, as no more than 10 percent of them reported taking any one class of trip. This mobility and apparent willingness of many snowmobilers in Region III to travel considerable "'"As was noted above, 28 percent of the snowmobil­ ing days originating in Region III was spent in the other two regions. 157 distances indicates that they not only respond to chang­ ing snow conditions but also exercise considerable selectivity in choosing sites for snowmobiling. ever, How­ for this kind of information to be useful to planners and managers, it should be based on much smaller geographical units. Recommendation.— It is recommended that high priority be given to the study of snowmobiling in the Lower Peninsula, in particular, those areas that contain heavily used resources and those areas where large numbers of snowmobilers' residences are concentrated. In such a study, the variables that should be given primary consideration are: (1) number of days for which there is a specified minimum number of inches of snow cover (which should provide a better indication of the snowmobiling potential of the area than does average snowfall), (2) classes of land used, including private lands, (3) types of activities pursued, (4) preferences of users in terms of the resources and facilities, and (5) travel patterns of snowmobilers. Implications for Planning and Management Need to examine changing behavior.— We are living in an era of rapidly changing behavior patterns. This phenomenon is readily observed in outdoor recreation where both new and old activities experience periods of 158 rapid growth followed by plateaus in participation example, snowmobiling and cross country skiing). (for Such changes in recreation behavior frequently impose pressures to change existing policies and management procedures. However, according to Chubb, there are only a few studies that appear to have the detection of change as a major goal.^ Apparently, many agencies view research primarily as a tool to help solve immediate problems— from improving management of existing resources and facilities to justifying continued financial support. In such a changing society it is very likely that such myopic studies will be out of date by the time they are com­ pleted. In addition to many studies being designed for a single purpose, the problem of examining change, as Chubb goes on to point out, is further complicated by study designs being changed from one study to another to 2 such an extent that comparisons of data are impossible. Such a situation makes it difficult to determine trends in behavior. Michael Chubb, "Recreation Behavior Studies: Emperical Indicators of Change" (paper presented at the National Research Symposium on Indicators of Change in the Recreation Environment, Pennsylvania State University, University Park, Pennsylvania, July 9, 1974), p. 10. ^Ibid., p. 12. 159 Recommendation.— It is recommended that new snow­ mobile studies be conducted in order to determine trends in snowmobiling behavior. In order to accomplish this task, such investigations should use some or all of the variables study. (with comparable structures) used in this These trends could then be compared to other known trends in recreation and provide land management agencies with a clearer indication of possible future changes in recreation desires and behavior. Need to forecast behavior.— In order to plan more effectively for the future, it is important for planners to be able to forecast use patterns from a knowledge of other variables, such as user characteristics, snowfall patterns, etc. Not only would this kind of information facilitate long-range planning, but it would enable agencies to anticipate changes in use patterns that effect their management programs. Recommendation.— It is strongly recommended that a study be undertaken to establish the functional relationship between selected user and resource char­ acteristics and the use patterns of snowmobilers. It is suggested that the AID algorithm be used to determine the nature of the interaction effects between the selected independent variables. Once these relationships have been established, an appropriate multiple regression 160 model that will accommodate these interaction effects should be developed to determine the set of variables that best predict snowmobile use. Application of Recreation Research Generally Life style of u s e r s .— Studying and determining the behavior patterns of participants in a particular activity, or even in several, can provide very useful information. However, according to Chappelle, in order to gain an understanding of the recreation phenomenon we need to broaden our scope and examine the total process.^ In other words, we need study the life styles of various user groups and then to assess the place that a particular outdoor activity plays in these life styles. If, as researchers we can reach the point where we have achieved significant insights into the role that outdoor recreation occupies, or the "needs" that it attempts to satisfy, then we will be much closer to being able to deal intelligently with the whole aspect of land use priorities and selection of alternative activities. For example, it would be useful to know if it is the outdoor environment or the social aspects or 1Daniel E. Chappelle, "The Need for Outdoor Recreation: An Economic Conundrum?" Journal of Leisure Research, 5 (Fall 1973) : 52. 161 some combination of both that is meeting these needs. It is very possible that many activities have a fairly low priority rating in some users' life styles, and, if forced to make a choice either of an economic or of an activity nature, would reject that activity as only a "frill" in his life style. Recommendation.— It is recommended that serious consideration be given to the development of research studies that will examine the life styles of outdoor recreators. It is further suggested that such an endeavor be conceived and designed in cooperation with other fields concerned with human behavior such as psy­ chology, sociology, and economics. Underlying values.— It frequently appears that when agencies decide upon an area of concern to be examined scientifically, they fail to ask a fundamental question such as: are they responsible for providing for such an activity and, if so, to what extent. Often, they seem to be responding to public and political pressures. Chappelle, in questioning the need for recreation, states that: " . . . while many recreation professionals appear to regard availability of recreational experience as an inherent right of all citizens, they seem also to be relatively unconcerned with providing 162 all citizens equal opportunities to experience recreation."'*' In addition, professionals are fre­ quently heard to expound on their mandate to protect the resources under their jurisdiction, yet they appear to encourage, or at least permit, environmentally damag­ ing motorized recreation activities within their boun­ daries. Is this attitude primarily one of responding to the vocal and affluent segment of society so that their operation will run more smoothly? These questions are not easily answered given the nature of the political decision-making process. Planners and managers are not the only ones who should consider these problems. Researchers must also ask themselves these questions throughout their work. As our natural resources become scarcer and if participation in outdoor recreation continues to increase, then these are questions that should be the concern of all those involved with recreation. ■*"Ibid. , p . 51. APPENDICES APPENDIX A QUESTIONNAIRE LETTER OF TRANSMITTAL AND MAP APPENDIX A MICHIGAN STATE U N I V E R S I T Y east lansing ■ M ichigan mhzs DEPARTMENT OP PARK AND RECREATION RESOURCES • NATURAL RESOURCES DUJLDINO r “ 'i Dear Snowmobiler: Now that the snowmobile season is over we would like to ask for a few minutes of your time to answer some questions concerning where and how much you used your snowmobile during the past season. We are conducting a survey of snowmobile use in Michigan in coopera­ tion with the Michigan Department of Natural Resources. The purpose of the study is to learn more about how snowmobiles are used and the characteristics and opinions of the users. This information will help us forecast the future growth of snowmobiling and facilitate the planning and development of special areas to meet the growing needs of Michigan snowmobile operators. Your name was selected at random from a list of all registered snow­ mobile owners in Michigan. We assure you that your answers will be held in strictest confidence. They will only be used with all other replies to show the patterns of snowmobiling and the opinions of snowmobilers in Michigan. Please assist us in carrying out this survey by completing the enclosed questionnaire, it should only take 15 to 20 minutes of your time. Sincerely, Michael Chubb, Director Recreation Research & Planning Unit MC:sjh 163 Norman F. Smith, Chief Recreation Resource Planning Division Michigan Department of Natural Resources 164 FOR YOUR ASSISTANCE: A COUNTY AND HIGHWAY MAP OF MICHIGAN IS L E R O Y A LE O O C E 0 IC M ARQUETTE LU C E SCHOOLCRAFT tSCAOAtA CHSBOrtAM EM M ET (TRESOUE jO TSE O O LAN A m o NTMOR. i * t- P E N A N T R IM |K A L K A S K A |CRAW F0Ro l O S C O D A HI IE gR i / ( AL C O N A 1 I***YlSm6 r __ 'T*_I_|____ ",,8Tf^f"r ,AHSSA«RCE| 1O S C O M .IO Q E M A W I IOSC O Jvjl 'J f yA,C*DILAC || Tl*«v - ro??or*rc:,ii roar Avsna tVOlrntTOH i t * : W AYOO 'M E C O S T A j SCOLA MUSKC. i .i.X' | M O N TCA LM KENT o r iOT I O R IA tT IO \ ’ t I I «l I ^'0N , C LIN TO N jS H IA W A . ( ■” —i“^LAR££Ri r - - - -t-— I- -r - - - 16 E N E S E E ' i 1 I BARRY ERTOH^li'fenHRM 'u v m w r i X CALHO UN KALAMAZOO CASS j| O AKLAND r* |S T . JO S EPH ! q p a NC H ) JA C K S O N ' JACKtO* I H IL L S D A L E L A IR roar auaoa FL/ar WASHTENAW | APPENDIX B REMINDER CARD APPENDIX B Dear S n o w m o b i l e r : This is a r eminder about the SNOW MO BIL E que st io nna ir e wh i c h was mai l e d to y o u recently. We have b e e n receiving completed q u est io nn ai res f r o m other s n o wm ob il e owners but at our last check yours was not among them. We realize that the q u e s t ion na ir e does take some time and effort to fill out. Ho wever it is important that we rece iv e your reply, not only for this study but also for p la nn in g future sn owmobile facilities. Please send us your c om pleted questionnaire! Th a n k you, M i c h i g a n D e p ar tm ent of N a t u r a l R es ources M i c h i g a n State U ni v e r s i t y 165 APPENDIX C REVISED LETTER OF TRANSMITTAL APPENDIX C M I C H I G A N DEPARTM ENT STATE OF PARK A N D UNIVERSITY R E C R E A T IO N EAST I.A N S IN G RESO UR CES • N A T U R A L • M IC H IG A N IMS 21 RESO UR CES M U IE D IN G De ar Snowmobiler: Several weeks ago we mailed to you a S N OWM OB IL E que st io nna ir e, asking for information about how you used your s n o wm obi le duri ng the past season. So far we have not heard from you. In case the q uestionnaire has gone astray, we have en cl os ed another wh i c h we hope you will complete and retu rn as soon as possible. This survey is rapidly coming to a close and it is important that your information be included in the survey results, so that a mor e ac c u r ­ ate picture of M ic hig an snowmobilers m a y be obtained. The information you supply will be greatly ap pr ec i a t e d and will, of course, be treated confidentially. Wou ld you p l e ase complete this q u e s ti on na ir e and return it in the e nv elope provided. Sincerely, rMichael Chubb, Director R e c r e a t i o n R e s e a r c h & Planning Unit N o rm an F. Smith, Chief R e c r e a t i o n R e s o u r c e P lan n i n g Di vi si on M ic h i g a n D e p a r t m e n t of Natu ra l Resou rc e 166 APPENDIX D TELEPHONE INTERVIEW SCHEDULES FOR NONRESPONDENTS APPENDIX D SITVEY n r ?:on-'n.Frpr»iTr'rr’TS TELEPHONE IETr'VIEt’ ror. r I1CHIGAI! SITOTTSILE USE STUTY County of Residence i 1 L_ Answer to call Contacted party Call back at yes I— 1 1 refusal accepted or Pate Time Do answer Busy signal □□ Interview □ no 0 □ □ Phone T'ronq >To. | . : . ; 000 000 | 0) O Poes not own a snowmobile Dumber not available iirrmaE’T Hello! Hay 1 speak to !‘r. (Krs.)_______________________________________________ . tiy name is________________________________ . I am ohonine on behalf of ’"Ichiyan Sate University and the Department of natural Resources about the Snowmobile Ouestionaire that was 6ent to you some time apo. (If they indicate that they have sent it in, then ----) 1. VJhen did you mail it? 2. Did you fill it out Before July 8 After July 8 no yes | | □ (I! so -- terminate) Vie are calling a number of people in your county "ho did not fill out a ouestionaire and v;ould appreciate it if we could take about ten minutes of your tine to ask you a few questions. I want to assure you that your answers will be treated confidentially. (Reaction) ________ _____________________________________________________________________ 1. Are you the HEAP of your household? yes If no; wife VJhat is your relationship friend 167 □ 0 0 no son other □ 1__1 daughter 0 (0 168 2. How, would you tell me the cake, horsepower, and owner of each snowmobile owned by members of your household? Horsepower i.'ake Registered owner, e.g., head, wife, son, etc. (Interviewer comments) 3. He would like to ask some questions about the counties you used for snowmobilinu. a) First, which county did you use the most for snowmobiling during thepast winter? How many days did you go snowmobiling there? b) (Count part day as awhole day.) tfhat was the county that you used the r. xt most? Number of days? c) The county you used the third most? Humber of days? d) How many days did you snowmobile in other counties? County of nost use County of 2nd most U3s County llama Ho. of days used (Interviewer comments) A. In what year did you buy your first snowmobile? County of 3rd most use All other countv use 169 5. ilov?, would you tell us the hind of land you used for snowmobiling by indicating which of the followin'? you used? (Place a checl- opposite the classification mentioned.) Kov many days did you r o smmcobiling there? n n n n n Your own land Federal land State owned land days Private land, pay fee n n day9 Local public roads not plowed days City nark land (including City| golf courses County owned land n > Private land, no charge i days Labes & rivers days Other days days days [ days n days (specify) (Interviewer comments) 6. In order to forecast future demand for snowmobile areas in Michigan, it i3 necessary for us to be able to relate family characteristics to use patterns of snowmobilers. He would appreciate your answers to these questions. a) I’hat is the age and sex of the HEAD of your household? b) '.That is the occupation of the HEAD of your household? c) How many grades ofschooling have youcompleted? 4 8 Age_____ /ale Female__ Occupation (not organization) 5 6 7 9 j 10 11 12 J high school d) Ue would 1 13 (Circle one). 14 15 university 1C J 17 or more post grad. ! liketo have some idea of the total Income of I your household. I will read off a number of income catorgories, please stop me when I reach the right one. under $3,000 f~) $8,000 - $9,999 □ $20,000 - $24,999 □ $3,000 - $5,999 n $10,000 - $14,999 D $25,000 - $29,999 Lj $6,000 - $7,999 n $15,000 - $19,999 □ $30,009 and over I 1 (Interviewer comments) _________________ _______________ 170 7. Uould you tell ne which of the following activities you or members of your household took part in while us in?, your snowmobile last season? How many days did you spend on this activity? (Count each part day on an activity as one day.) Scrambling in open areas and on lakes . ( I Snowmobiling to work or during your work D. Competive racing _days days Trail riding and forest cruising days Other___________________ (specify) jdays (Interviewer comments) 8. We are very interested in learning what reasons you had for not filling out the questionaire. Would you care to comment? 73 Please feel free to be quite frank. Thank you for your cooperation and assistance with this study. APPENDIX E SNOWFALL INDEX AND AVERAGE INCHES OF SNOWFALL APPENDIX E SNOWFALL INDEX r~“ . SCNDOLCNAFY '■ > MACKINAC 4 D ELTA tM M C T | IfNisguc ^ \ C M A N lf V O I X L L - A - H C - “ - jo m e o A N T ftiM IBONTMON?! N '- P C M r I v/ |__ Range l/(/!ltAL,,AS*AlcNAWfOND I O IC O D A | ALCOAA rSCN7lC'ggO'THAV, Cl ass ■/ i o I i 6 ^ 5 ^KitAUKlil*0,C0“- OAEMAtf~ “ l i i e o 30-44 inches 1 45-59 inches 2 60-74 inches 3 75 -8 9 inches 4 7_! 6 1 . 4 _L_4_J_3_ _L. ---UAPE - ^ T LAKE ” r««F.r. 1 I I I J1AAIAAC p ' 4 ! 3 : 2 ! 2 f.h/~ \ 7_:_4_j _3_;_ 2_J_2_ H ,/o CIAIIa T i« » ‘*»° ,MECO»TA|'**,tLL*|lllOLA»s', n V 6.: 4 : ?.!.! \y u 9 > f.> 90- 10 4 inches 5 I I \\ ly O A IC A L P I 1 < --- — *■ t OTTAWA | ~ I l<,N' 4 II ---------------- 1 105-119 inches 6 AVLAAAI. 120 inches and over 7 , uu‘ jvA*. .aMon. t , ;«A « A AT TI IOOITI I * * • ' * * » O I1 O2 1. 3 2 1 , 1 _ ,_____ r , | IA T 0 . IT| r--'-p I I :i 3 ; 2 ;i ^ 1 I __ I______________________ [ f k a llaamm aa ii..^* c a l m o u m | 2 I O j____» '__T « , i j OA NLAM b W lP M T O I, a m a f>A COmt I , : i I :1 _ — — T ( ja c k o o n 1 ! i C t l A T OAA,*"1 l * " '1* ® * ’ I > -3 1 p , O I £. — — — WAVME 'I w , a in t in a w j f " w 1 - 11 \ r I 172 AV E RAG E INCHES OF S NO W FALL WINTER OF 1969 - 70 UllAMAo INn IC4ATO 136 | 15 I *AANfll|| 111 -IB I " 1 1 MICHIGAN DEPARTMENT OF STATE HIGHWAYS W j 0 I *C*AtCO ( MlCOtlA j I LOCAL GOVERNMENT DIVISION LEGEND 8? AVERAGE INCHES SNOWFALL 14 CHANGE FROM LAST YEAR ,DNj Ml L IMGNKAIM 1G*Afio? T r _| J . _ 53 72 j AlU O N +9 oni I 02 1 46 ~"| ONION j J 03 *OMA 63 IAJW I « ! j IAION I 04 I NGHAM m T „ , . I ClAJ» HWAWMhTI 36 j t24 1 lIVNGJtON I <4 ±:"-L'2 l L I j *13 )_ _ _ L 51 59 T "j 31 l^.-L , i a. I j | 06 42 APPENDIX F CLASSIFICATION OF VARIABLES APPENDIX F TABLE 33 CL AS S I F I C A T I O N OF VARIABLES Scale Q uantitative Variables Age of Household Head E d uc a t i on Level Combined Gross ratio3 of Household Head Income of Number of C h i l d r e n , Household 18 y e a r s and Under , i n Household Horsepower of Snowmobile Year of First ratio3 i nterval^ ratio ratio Snowmobile Purchase i nterva1 Number of Snowmobiles Owned by Household ratio Number of Snowmobi l ing Days rati o Number of Days on Each Cl a s s o f Load ratio Number of Days f o r Each Snowmobi l ing A c t i v i t y ratio Number of Days f o r Each A c t i v i t y A s s o c i a t e d Wi t h Snowmobil ing ratio Number of O v e r n i g h t T r i p s Qualitative ratio Variables Membership i n Snowmobi ling Clubs nomi nal Opi ni on Toward P r e s e n t Snowmobi l ing R e g u l a t i o n s o r d i nal Opi n i on Toward Enf or cement of Snowmobi ling Regulations o r d i nal 173 174 Qualitative Scale Variables Op i n i o n Toward t he O p e r a t i o n o f Snowmobiles by C h i l d r e n 14 y e a r s of Age and Under nomi nal Op i n i o n Toward R e g u l a t i o n s Governi ng Snowmobile Noi se nomi nal Op i n i o n Toward R e g u l a t i o n s Governi ng Snowmobile A c t i v i t y Near I c e - F i s h i n g nomi nal Opi n i on Toward A l l o w i n g Snowmobi l ing on P u b l i c Thoroughfares aThe s t a t i s t i c a l a n a l y s i s was c a l c u l a t e d t he o b s e r v a t i o n s were gr ouped. nominal prior to ^Income i n f o r m a t i o n was c o l l e c t e d i n ni n e c a t e g o r i e s . For s t a t i s t i c a l purposes i t was assumed t h a t t hese were equal a p p e a r i n g i n t e r v a l s . APPENDIX G WEIGHTS USED IN EXPANSION OF DATA APPENDIX G W E I G H TS FOR 11 SUB-REGIONS. Region I 02 07 17 21 22 27 31 36 Alger Baraga Ch i ppewa Delta Dickinson Gogebic Houghton Iron 35 42 48 49 52 55 66 75 Keweenaw Luce Mackinac M a r quette Menominee O n t o nagon Schoolcraft ~1 45 51 53 54 56 57 60 62 64 65 67 68 69 71 72 83 Le e l a n a u — Manistee Ma s o n Mecosta Mi d l a n d M i s s aukee Montmorency Newaygo Ocea n a Ogemaw Osceola Oscoda Otsego Pr e s q u e Isle Roscoimnon Wexford 44 46 47 50 58 59 61 63 70 73 74 Lapeer Lenawee Livi n g s t o n Macomb Monroe Montcalm Muskegon Oakland Ottawa S aginaw Sanilac Shiawassee St. Clair St. Joseph Tuscola Van Buren Washtenaw W ay n e 35 17 ~j 35 J R e g i o n II 01 04 05 06 09 10 15 16 18 20 24 26 28 35 37 40 43 Alcona Alpena 41 Antrim Arenac Bay 22 Benzie Ch a rlevoix Cheboygan Clare 41 Crawford Emmet Gladwin Grand T r a v e r s e - 16 Iosco Isabella 41 Ka l kaska Lake 41 R e gion III 03 08 11 12 13 14 19 23 25 29 30 32 33 34 38 39 41 Allegan ~ Barry Barrien Branch Calhoun Cass C linton Eaton _ Genesee Gratiot ~ Hillsdale Huron _ Ingham Ionia — Jackson Kalamazoo _ Kent - 56 36 56 76 77 78 79 80 81 82 20 56 23 175 56 33 56 25 APPENDIX H MICHIGAN'S PUBLIC LANDS STATE GAME Petobego 49 Tobico Marsh 36 Lost Nation 50 Mcrtiny Lake * * 39 Pomte MouiHee 5i Tuscola Vasscr Middievilie 40 r'Ort Huron 52 vestcburg 41 Portland 26 Fulton 27 GourdnecK 28 Grand Haven 29 M m a e n City Gratiot • S o g m a w 30 Murphy Lake 42 Rogue River Gregory Haymorsh 31 Muskegon Rusn L c ^ 32 Oak Grove 43 44 Lanqston 33 Lapeer 34 Onsted Pentwater River Leidy Lake'' 35 36 I £ S T A T E W H D l IFE A R E A S 1 I Lowell Manistee Rive' 1 1 25 Flat River II Deford HOUGHTON 3? Erie Cass City KEWEENAW AREAS Petersburg state wildlife 46 47 48 Three Rivers p e s e a r c Wildfowl 3 c 61 Houghton Lake 59 Unadilla 62 Beaver islands 63 Rose Lake 54 Ncyanqumg Pom* 55 Guamcassee 56 St Clcir Fla? - AREAS STATE G A M E "AR7 6a V.:so'' 56 57 cC-rs$' Shiawassee R-ve' S*an*on O0 U?TC Fish Pomr c C^ppewa 4 f. Shcrcovi'ie W i g w o m Bay 53 ■h Sen, : c c 7 ** A Sanilac 176 ONTONAGO N" MARQUETTE® G O G E B IC m aw uett CHIPPEWA WO©muaft ; CRYSTAL PfiLLSi 2«CKiNSON MACKINAC H i NOMINE! PRESQUE ISLE C HARLEVOIX A LPE N A ANTRIM E E IA N A U 37 GRANO M IC H IG AN •E N Z lc 1TKAYE2SE M A N flT E E - W c X rO K D ALCO NA KA LK ASK A D E P A R T M E N T OF N A T U R A L RESOURCES i MICHIGAN'S PUBLIC LANDS MISSAUKEE ROSCOMMON R O SC O M M O N 6 iW ' & 2 OGEMAW io s c a MARQUETTE(@J\^ -GOGEBIC1 MARQUETT^ NEWBERRT © CHIPPEWA C-HrSTflL PALlsI MACKINAC ICKiNSON MEMOtUNCE EM M ET CHEBOYGAN PRESQUE ISLE CHARLEVOIX 0 ^ MONTMORENCY i ’F> OTSEGO .iY^OSD GRAND BENZIE irtAYEtes M AN IS T E E WtXFOSD MICHIG AN KA LK ASK A ALPE NA OSCODA CRAWE 08 D ALCO NA ^ M.O D E P A R TM E N T OF N A T U R A L RESOURCES mROSCOMMON g . MICHIGAN'S PUBLIC LANDS 611^OSCO2SMON MISSAUKEE CADiLLAv© - CLARE ; N A T IO N A L F O R E S T S ,P A R K S , A N D REFUG ES S T A T E FORESTS STATE G A M E ,W IL D L IF E ,A N D R E S EA R C H A R E A S m- ^ 34 A OCEANA IS ABELLA MECOSTA N E W A jG O H U RO N v* 58 BAY 154 **53 27 176 OSCEOLA io s c a ARENAC G L A D W IN M ASO N - OGEMAW 49V^55 M ID LA N D . ‘l.J- S TATE R E C R E A T IO N A L A R E A S S A N ILA C i m g n tc a lI? STATE PAR KS SA@ R (TSSt ) and N± > M (1.3.2) where R is an arbitrary parameter (normally .01 < R < .10) and M is an arbitrary integer (normally 20 < S < 40). The requirement (1.3.2) is made to prevent groups with little variation in them, or small numbers of observations, or both, from being split. That group with the largest total sum of squares (around its own mean) is selected, provided that this quantity is larger than a specified fraction of the original total sum of squares (around the grand mean), and that this group contains more than some minimum number of cases (so that any further splits will be credible and have some sampling stability as well as reducing the error variance in the sample). 3. Find the division of the classes of any single predictor such that combining classes to form the partition p of this group i into two nonoverlapping subgroups on this basis provides the largest reduction in the unexplained sum of squares. Thus, choose a partition so as to maximize the expression (n^i where and + n2y2 ) = n^ + 1*2 Yt = "1*1 + (1-2-3) n2 h 177 178 for group i over all possible binary splits on all predictors, with restrictions that (a) the classes of each predictor are ordered into descending sequence, using their means as a key and (b) observations belonging to classes which are not contiguous (after sorting) are not placed together in one of the new groups to be formed. Restriction (a) may be removed, by option, for any predictor X^. 4. For a partition p on variable k over group i to take place after the completion of step 3, it is required that BSSikp ^ Q (1.3.4) where Q is an arbitrary parameter in the range .001 < Q < R, and TSS,j, is the total sum of squares for the input sample. Otherwise group i is not capable of being split; that is, no variable is "useful" in reducing the predictive error in this group. The next most promising group (TSSj = maximum) is selected via step 2 and step 3 is then applied to it, etc. 5. If there are no more unsplit groups such that requirement (1.3.2) is met, or if, for those groups meeting it, requirement (1.3.4) is not met (i.e., there is no "useful" predictor), or if the number of currently unsplit groups exceeds a specified input parameter, the process terminates. AID (2) ALGORITHM Preliminary Read in. Steps 1 and 2. Read in all parameters and all input observations, including all predictors and the dependent variable Y. Screen out observations where Y is missing data or it is not desired to use this observa­ tion. Save all observations on tape if necessary. To start, identify all observations used in the analysis as belong­ ing to group number one. Group number one is the current candidate group. Go to Step 6. Test for Termination of the Procedure. Step 3. Determine whether or not the current number of unsplit groups is about to exceed the maximum permissible number; if so, go to Step 22, as the problem cannot proceed further. Determine Which Group Should Be Selected for Attempted Partitioning. Steps 4-6. Considering all groups constructed so far, find one of them such that a. the total sumof squares (TSS^) of that or equal to R per cent of the total sum input observations (TSS^); group is greater than of squares for the b. the number of observations in the group is not smaller than MSIZE; c. the group has not already been split up d. there has been no previous failure to split up the group; e. the total sum of squares of that group is not smaller than the sum of squares for any other group that meets the above four criteria. into two other groups; If there is no such group, go to Step 23; the problem is complete. The group selected is the current candidate group, which' will be the subject of an attempted split.- Identify it with its group number (i) and print out Niy ZY± , ZY^, and T S S ^ 180 Partition Scan Over All Predictors. 7. Set 8. Increment j by 1. If j is larger than;the number of predictors being used in the analysis, the partition scan is complete; go to Step 20. 9. j = 1 Steps 7-19. and go to Step 9. 2 Compute N . . , Z Y . . , Z Y . . ,Y . . ^ ijc’ ijc* rjc’ i]c over group i. for each class c of predictor j r J 10. Determine whether or not there exist two or more classes c, such that % j c 4 0. If not, predictor j is aconstant over group i; print an appropriate comment and go to step 8. 11. If predictor j has been defined as monotonic, skip Step 12 , do not sort the Step 9 statistics, go ta Step 13 instead. 12. Sort the statistics produced in Step 9, together with the class identifiers for predictor j, intodescending sequence using Y.j. as a key. Partition Scan Over the c Classes of Predictor j. 13. Steps 13-17. Set p = 1 and go to Step .15. 14. Increase p by 1. If p is .larger, than .(c.- - 1) , where c. is- the number of classes in the jLth predictor,?then print the statistics for class Cj and go to Step 18 as all possible feasible splits have been examined. 15. If ZEI^ = = 0 for k — 1, ... p,mr. if ' (N^- - N^) = N 2 = 0, got. to Step 14 as this split cannot b e made because of empty classes, in this group for predictor j. Otherwise, compute BSSp , the betweengtoups sum of squares for the attempted binary split of group, i on predictor j between the sorted classes (1, ...., p) and the adjacent sorted classes (p^K 1,. .. . . c) . Print the statistics, for class p. 16. If this BSS_ is not larger than any BSS previously computed: for this predictor over this group, go.-to p. Step 14. 17. *fiiis is the largest BSSp encountered? so- far.: for this predictor. Benember BSSp and the partition number p ^ print them and ga to Step 14. 181 Determination of Best Predictor. Steps 18-19. *18. Was the maximum BSSp for predictor j larger than the largest BSSp obtained from any or the other predictors previously tested over group i? If not, go to Step 8. 19. This is the best BSSp produced by any of the predictors tested so far over group i. Remember this partition and this predictor and then go to Step 8. Is the Best Predictor Worth Using? *20. 21. Steps 20-21. Was the maximum BSS retained after the scan of all predictors over group i equal to at least Q per cent of the total sum of squares? If not, mark group i as having failed in a split attempt and then go to Step 4. Group i is to be split into two new groups and destroyed. Using the class identifiers and the partition rule remembered from Step 19, split the observations in group i into two parts. Identify the two new groups as having been created. Identify group i as having been split. Print the statistics from the successful partition attempt. Increase the total number of groups created so far by the quantity 2. Increase the current number of unsplit groups by one. Then go to Step 3. Termination of the Algorithm. Steps 22-26. 22. The maximum number of permissible unsplit groups has been reached. Print an appropriate comment and go to Step 24. 23. There are no more groups eligible for further splitting. an appropriate comment and to go Step 24. 24. Print out a summary record of all groups created in the process of splitting, including the group number, its parent group, the values of the predictor class identifiers that were used in the partition which constructed the group, the predictor number used in this partition, an indication of whether or not this present group was ever split, and , ZY^, and TSS^. 25. Determine whether punched or tape residuals are desired. go to Step 26, otherwise go to Step 1. 26. Compute predicted values of Y and residuals and, by option, punch them and/or write them on tape with the data. Then go to Step 1. Print If so, *These decision rules constitute the crucial steps in the algorithm. 182 Formulas Y ZY/N TSS ,t BSS v v & Yi ) 2 n + % o 2 — N2 Y a « Y ‘i (z y ) M CO R a WSS Source: ** Y a — Y a * TSS - BSS Sonquist and Morgan, Detection of Interaction Effects, pp. 5-6, 158-61. APPENDIX J W E I G H T S U S E D IN A I D P R O G R A M APPENDIX J WEIGHTS USED IN AID PROGRAM * Source of Data Region I Marquette County A ll Other Counties Region I I Bay County Grand Traverse County A ll Other Counties Region I I I Genesee County Ingham County Kent County Oakland County Wayne County A ll Other Counties Weight 331 669 TOSO 174 159 667 T000 125 96 114 120 106 440 Tooo Each weight is the proportion (times 1,000) that the number of observations in a sub-region is of the number of observations in that region. APPENDIX K P R O P O R T I O N OF V A R I A T I O N IN N U M B E R OF S N O W M O B I L I N G DAY S E X P L A I N E D B Y P R E D I C T O R V A R I A B L E S APPENDIX K TABLE 34 PROPORTION OF VARIATION IN NUMBER OF SNOWMOBILING DAYS EXPLAINED BY PREDICTOR VARIABLES Pred i c t o r Variable P r o p o r ti o n of V a r i a t i o n ( b s s d/ t s s t ) Explained Region Region Age of Household Head .0354 Oc c u p a t i o n of Head . 0291 I Region I I .0375 Household E d u c a t i o n Level o f Household Head .0110 .01 53 . 0241 . 00 6 8 Income o f Household .0121 . 01 0 5 Age Range of C h i l d r e n . 0484 . 0061 Horsepower .0130 Ownershi p of Snowmobile .0135 . 0093 Number of Snowmobiles Owned by Househol d . 0 337 . 06 4 4 Years of Snowmobi ling .0070 Depth of S n o w f a l 1 . 0 277 R e g u l a t i o n s Near I c e - F i s h i ng . 0234 Membership i n Snowmobile Cl ub .01 16 T o t a l P r o p o r t i o n of V a r i a t i o n Explained ( bss/ t s s ) t .0341 .140 184 . 208 .0301 .1 58 III SELECTED BIBLIOGRAPHY SELECTED BIBLIOGRAPHY Books and Reports Baldwin, Malcolm F. The Off-Road Vehicle and Environ­ mental Quality^ Washington, D.C.: The Conservation Foundation, 1970. Burton, T. L . , and Cherry, G. E. Social Research Tech­ niques for Planners. London: George Allan and Unwin Ltd., 1970. Cochran, William G. Sampling Techniques. John Wiley and Sons Inc., 1963. New York: Crapo, Douglas, and Chubb, Michael. Recreation Area Day-Use Investigation Techniques: Part I, A Study of Survey Methodology. East Lansing, Michigan: Recreation Research and Planning Unit, Michigan State University, 1969. Chubb, Michael. Outdoor Recreation Planning in Michigan by a Systems Analysis Approach: Part III— The Practical Application of "Program RECSYS*1 and "SYMAP7" East Lansing, Michigan: Recreation Research and Planning Unit, Michigan State Uni­ versity, 1967. Directional Marketing Company. 1970 Snowmobiler Survey. Duluth: Upper Great Lakes Regional Commission, 1971. Driver, B. L . , and Tocher, S. Ross. "Toward a Behavioral Interpretation of Recreational Engagements, with Implications for Planning." Elements of Outdoor Recreation Planning. Edited by B. L. Driver. Ann Arbor: University Microfilms, 1970. Hays, William L. Statistics. Toronto: and Winston Inc., 1963. Kish, Leslie. Survey Sampling. and Sons I n c ., 1965. 185 Holt, Reinhart New York: John Wiley 186 Klopchic, Peter. An Analysis of Snowmobiling in Ontario/ Winter 1969-70. Toronto: Department of Tourism and Information, 1971. Maxwell, A. E. Analyzing Qualitative D a t a . Methuen and C o . L t d ., 1961. London: Minnesota Department of Conservation, Bureau of Planning. Minnesota Snowmobile Study, 1970. St. Paul: Minnesota Department of Conservation, 1970. Moser, C. A. Survey Methods in Social Investigation. London! Heinemann Education Books Ltd., 1958. Oppenheim, A. W. Questionnaire Design and Attitude Measurements. New York: Basic Books Inc., 1966. Selltiz, Clair; Jahoda, Marie; Deutsch, Morton; and Cook, Stewart W. Research Methods in Social Relations. New Y o r k : Holt, Reinhart and Winston, 1959. Sonquist, John A. Multivariate Model Building. Ann Arbor: Institute for Social Research, University of Michigan, 1970. ________ , and Morgan, James N. The Detection of Inter­ action Effects. Ann A r b o r : Institute for Social Research, University of Michigan, 1964. Articles and Proceedings Briggs, Steve F. II. "A Look at Where Industry Is Going." Proceedings of the International Snowmobile Con­ ference! Albany, N.Y.: New York State Conservation Commission, May, 1969. Chappelle, Daniel E. "The Need for Outdoor Recreation: An Economic Conundrum?" Journal of Leisure Research, 5 (Fall 1973) . Dodge, Robert O. "Michigan." December, 1970. Parks and Recreation, Glasgow, Leslie L. "Snowmobiles Today." Trends in Parks and Recreation, 6 (October 1969) . Herrmann, Robert O. "Interaction Effects and the Analysis of Household Food Expenditures." Journal of Farm Economies, 49 (November 1967). 187 Howe, Harold K. "Industry's View of Problems that Face Snowmobiles." Proceedings of the International Snowmobile Conference. Albany, N.Y.: New York State Conservation Commission, May, 1969. Koenings, Roman H. "Introduction." Proceedings of the International Snowmobile Conference. Albany, N.Y. State Conservation Commission, May, 1969. Market Research Report. "1970 Snow Goer Consumer Survey." Snow Goer Trade Journal, July, 1970. Morgan, James N., and Sonquist, John A. "Problems in the Analysis of Survey Data, and a Proposal." Journal of the American Statistical Association, 58 (June 1963). Van Doren, Carlton S., and Lentnek, Barry. "Activity Specialization Among Ohio's Recreation Boaters." Journal of Leisure Research, 1 (Autumn 1969). Public Documents Michigan. "An Act to Register and Regulate Snowmobiles." 74 Legislature, Regular Session, H.B. 3575, 1968. Michigan Department of Conservation. Snowmobile Demon­ stration A r e a s . East Lansing: Department of Conservation, 1968. Michigan Department of Natural Resources. Michigan Recreation Plan, 1970. East Lansing: Michigan Department of Natural Resources, 19 70. Michigan Department of State Highways. "Snowfall Contour Map, Winter of 1969-70." East Lansing: Department of State Highways, 1970. _________. Local Government Division. "Snowfall Data for 1970." Lansing: Department of State High­ ways, 1970. U.S. Department of Commerce, Bureau of Census. 1970 Census of Population— Michigan. Washington, D . C . : Government Printing Office, 1971. U.S. Department of Commerce, Environmental Data Service. "Climatological Data— Michigan, October 1969 July 1970." Washington, D.C.: Government Print­ ing Service, 1970. 188 Other Sources Chubb, Michael. "Recreation Behavior Studies: Imperial Indicators of Change." Proceedings of the National Research Symposium on Indicators of Change in the Recreation Environment. Pennsylvania State University,_ University Park, Pennsylvania, 1974. Emmery, Paul. "Try Hard." East Lansing: Urban Survey Research Unit, Michigan State University, 19 70. (Mimeographed.) Evinrude Motors. "Evinrude Snowmobile Trail Survey, 1967." Milwaukee, Wisconsin: Outboard Marine Corporation, 1967. (Mimeographed.) Igo, Alison. "An Analysis of the Validity of Mail Surveys for Use in Recreation Research." Master's thesis, Michigan State University, 1971. Lanier, L. L. "Snowmobile Survey of Selected State Parks." Unpublished material, Recreation Research and Planning Unit, Michigan State University, 1970. (Mimeographed.) Lininger, Charles A., and Warwick, Donald P. "Intro­ duction to Survey Research." Ann Arbor: Survey Research Center, Institute for Social Research, University of Michigan, 1967. (Mimeographed.)