.\ .m r. .‘q‘ RECREATION RESEARCH MAIL SURVEY TECHNIQUES: EFFECTS OF SELF-ADMINISTRATION AND NONI-RESPONSE Thesis for the Degree of M‘ S. MICHIGAN STATE UNIVERSITY ALISON JEAN CLINTON ISO 197 1 IIIIIIIIIIIIIIII I III \IIII ‘ 3 1293 01008 5987 MAR232001 1023 00 11190] ABSTRACT RECREATION RESEARCH MAIL SURVEY TECHNIQUES: EFFECTS OF SELF-ADMINISTRATION AND NON-RESPONSE BY Alison Jean Clinton Igo Natural resource land-managing agencies are becom- ing engaged in social science research as a means of assessing people's recreation needs, preferences, and behavior. Program develOpment, funding, and resource allocation are based on the outcome of these studies, so it is important that their results be reliable. One of the most pOpular techniques used in recreation research is the mail survey. In using this method to pre- dict future recreation trends, however, it is important to be aware of the effect which non-response may have on the validity of mail questionnaire results. The purpose of this study was to examine the problem of non-response in two recreation surveys, the 1968 Boating Demand Study and the 1970 Snowmobile Study, both under the direction of the Recreation Research and Planning Unit, Department of Park and Recreation Resources at Michigan State University. I 5.-.--. n.1,: TI questionna an non-r from thes compared turns, tc in each . level, t ticipatf use. D Statist any th used tc by eacj accept diffeI eithe] “01311. respo Sente Alison Jean Clinton Igo These studies were implemented by means of a mail questionnaire and follow-up interviews of both respondents and non-respondents in selected counties. Taking the data from these response groups, respondent interview data were compared with data from respondent interviewee mail re- turns, total mail returns, and non-respondent interviews in each county where a follow-up was done. Comparisons were made on the basis of educational level, total family income, amount of recreational par- ticipation, and geographic location of boat or snowmobile use. Data on the first three variables were compared statistically, yielding no significant difference between any two response categories. Descriptive statistics were used to illustrate the geographical distribution of use by each group in Michigan counties. 0n the basis of study results, it is possible to accept the hypothesis that there is no significant difference between respondents and non-respondents to either the 1968 Boating Demand Study or the 1970 Snow- mobile Study. Therefore, predictions based on partial response to each can be assumed to provide a valid repre- sentation of the needs, preferences, and behavior of a given recreation population. RECREATION RESEARCH MAIL SURVEY TECHNIQUES: EFFECTS OF SELF-ADMINISTRATION AND NON-RESPONSE BY Alison Jean Clinton Igo A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Park and Recreation Resources 1971 out this Michael under wI PIGtinq Providi Louis 1 and ad‘ Of my wOrkec‘ Ronalc thank ACKNOWLEDGMENT S I am indebted to many people who helped me through- out this project. Grateful recognition is given Dr. Michael Chubb, the chairman of my master's committee and under whom I worked as a research assistant while com- pleting my graduate work. The directors of the studies providing data for this project, Mr. Paul Fiske and Mr. Louis Lanier, are also to be thanked for their information and advice. Appreciation is also extended to the other members of my committee, Professor Sherilyn Zeigler, under whom I worked for a year; Professor Louis F. Twardzik; and Mr. Ronald Hodgson. Mr. Paul Risk, instructor, is also given thanks for his friendship and encouragement. Computer programming and statistical advice from Mr. James Mullin and Dr. Dennis Gilliland are gratefully acknowledged. The Michigan Department of Natural Resources supported the 1968 Boating Demand Study and partially sup- ported the 1970 Snowmobile Study; the use of data from these projects is greatly appreciated. ii I: help of 1 Anne Mil. and Mrs . and supg man-y th: Deadlines could never have been met without the help of the unit staff, particularly Randalyn Godley and Anne Mills, and many thanks go to them. A great deal of gratitude goes to my parents, Mr. and Mrs. Harry Igo, who have given continued encouragement and support throughout my education, and who have made so many things possible. iii Chapter 119 III Chapter I. II. III. TABLE OF CONTENTS INTRODUCTION AND PROBLEM STATEMENT . . . . Introduction. . . . . . . . . . User Surveys as a Recreation Research Tool. Mail Survey Technique. . . . . . . . Non-response in Recreation Research Mail Surveys. . . . . . . . . . Problem Statement and Objectives . . . . Assumptions . . . . . . . . . . . Studies to Be Examined . . . . . . . REVIEW OF RELEVANT STUDIES. . . . . . . Findings of Respondent-Nonrespondent Comparisons . . . . . . . . . . How Non-respondents Differ . . . . . Some Findings Suggest No Difference . . Methods Used in Other Studies . . . . . Summary . . . . . . . . . . . . STUDIES SUPPLYING DATA FOR COMPARISON . . . 1968 Boating Demand Study . . . . . . 1968 Study Questionnaire Design. . . . Sample Selection. . . . . . . . . Study Implementation . . . . . . . 1970 Snowmobile Study. . . . . . . . Snowmobile Questionnaire Design. . . . Sampling Procedure . . . . . . . . Study Implementation . . . . . . . iv Page (AUDI-4 ooumm 10 11 12 14 16 18 21 21 22 23 24 26 27 28 28 Chapter Page IV. ANALYSIS OF DIFFERENCES BETWEEN RESPONSE CATEGORIES O O I O O O O O O I O O 3 1 Characteristics Chosen for Comparison . . 31 Socio-economic Data . . . . . . . 31 Degree and Patterns of Participation. . 32 Response Categories Chosen for Comparison. 33 Format and Use of Questions . . . . . 35 1970 Snowmobile Study. . . . . . . 35 1968 Boating Demand Study . . . . . 37 Hypotheses . . . . . . . . . . . 41 Respondent Interview versus Respondent Interviewee Mail Returns . . . 42 Respondent Interviews versus Total Mail Returns. . . . . . . . . . . 44 Respondent Interviews versus Non- respondent Interviews . . . . . . 46 Techniques for Comparing Response Groups . 48 Education. . . . . . . . . . . 51 Income. . . . . . . . 52 Days of Boating or Snowmobiling . . . 53 Geographical Distribution . . . . . 54 V. RESULTS 0 O O I O O O O O O O O O 57 Education . . . . . . . . . . . 57 1968 Boating Demand Study . . . . . 57 1970 Snowmobile Study. . . . . . . 59 smary O O O O O O I O O O O 59 Income . . . . . . . . . . . . 6l 1968 Boating Demand Study . . . . . 61 1970 Snowmobile Study. . . . . . . 64 Summary . . . . . . . . . . . 65 Chapter VI. SELECT Chapter Amount of Recreational Participation 1968 Boating Demand Study . . . 1970 Snowmobile Study . . Summary . . . . . . . . . Geographical Distribution of Participation . . . . . . . VI 0 CONCLUSIONS 0 O I O O O O O 0 Problems and Limitations of Study . Interpretation of Results. . . . Implications for Future Studies. . SELECTED BIBLIOGRAPHY . . . . . . . . vi Page 67 67 69 69 72 81 81 82 84 87 Table Ma Ma Table 1. LIST OF TABLES Mail Returns and Interviews by County, 1968 Boating Demand Study . . . . . . . Mail Returns and Interviews by County, 1970 Snowmobile Study . . . . . . . . Comparison of Educational Level of Various Response Categories, 1968 Boating Demand Study 0 O O O O O O I O O I O O 0 Comparison of Educational Level of Various Response Categories, 1970 Snowmobile Study . Percentage of Survey Subjects from Different Response Groups in Various Income Categories, 1968 Boating Demand Study . . . . . . . Comparison of Total Family Incomes of Various Response Groups, 1968 Boating Demand Study . Comparison of Total Family Incomes of Various Response Groups, 1970 Snowmobile Study. . . Comparison of Amount of Boating of Various Response Groups, 1968 Boating Demand Study . Comparison of Amount of Snowmobiling of Various Response Groups, 1970 Snowmobile Study 0 O O I O O O O O O O O O 0 vii Page 26 30 58 60 62 64 66 68 70 Figure LIST OF FIGURES Figure Page 1. Comparison of Geographic Location of Snowmobile Use in Michigan Counties, Respondent Interviews versus Respondent Interviewee Mail Returns, Ingham County. . 74 2. Comparison of Geographic Location of Snowmobile Use in Michigan Counties, Respondent Interviews versus Total Mail Returns, Ingham County . . . . . . . 75 3. Comparison of Geographic Location of Snowmobile Use in Michigan Counties, Respondent Interviews versus Non-respondent Interviews, Ingham County . . . . . . 76 4. Comparison of Geographic Location of Snowmobile Use in Michigan Counties, Respondent Interviews versus Respondent Interviewee Mail Returns, Kent County . . 77 5. Comparison of Geographic Location of Snowmobile Use in Michigan Counties, Respondent Interviews versus Total Mail Returns, Kent County . . . . . . . . 78 6. Comparison of Geographic Location of Snowmobile Use in Michigan Counties, Respondent Interviews versus Non-respondent Interviews, Kent County . . . . . . . 79 viii CHAPTER I INTRODUCTION AND PROBLEM STATEMENT Introduction Many natural resource land-managing agencies are becoming engaged in social science research. Realizing the potential importance of being able to quantify people's activities, interests, and attitudes, in- vestigators have used results of these studies as a basis for establishing priorities and allocating millions of dollars and valuable natural resources. Crucial decisions are made on the basis of survey results. For this reason, it is vitally important to insure that predictions based on them are as accurate as possible. Affecting the accuracy and reliability of social research are forms of bias not encountered in physical or biological research. Trees make easy survey subjects because they are stationary and their character— istics are quantifiable. A person, however, equipped with reasoning power and distinct personality and psycho- logical traits, has within his power the right to deter- mine whether or not he will answer personal questions asked of him in a survey. undert. problez differ ticula tained the en predic lead t non-r. QIOUp Admin formu SOUrc time SOCie reCrE ine 1 to a] mandI Many Are Whenever a social science research project is undertaken, those who choose not to participate pose a problem. Do these people, known as "non-respondents," differ importantly from those who gig respond to a par- ticular survey? If they do, profiles based on data ob- tained from the latter will not present a true picture of the entire p0pulation being investigated. Consequently, predictions based on study results may be inaccurate and lead to a misallocation of funds and natural resources. It is essential, therefore, that the problem of non-response be examined by any discipline, agency, or group making use of survey research data. One type of administrator who relies heavily on survey results in formulating policies and programs is the recreation re- source decision maker. Charged with providing leisure- time activities for an increasingly leisure-oriented society, recreation planners must be able to assess recreation needs and desires and make predictions regard- ing future recreation requirements of the public. Faced with this challenge, planners are anxious to apply social science research techniques to the management of participation in recreational activities. Many of the tools available are sophisticated and refined. Are they reliable? One aspect of this question will be examined in this thesis; specifically, is non-response to recreation surveys a serious problem? F to them a recreatiI They can lating s pation 1' 0n the I demand evaluat such as Thirdly PIOblex it the CODdUC User Surveys as a Recreation Research Tool Recreation researchers have various avenues open to them as they investigate the types and amount of recreation activity in which people are participating. They can use the findings of other researchers, corre- lating such variables as income and recreation partici- pation in the past and projecting future participation on the basis of estimates of future income levels. The demand for recreation can also be measured indirectly, by evaluating sales of recreation goods and "by-products," such as boat gasoline, fishing supplies and the like. Thirdly, investigators may take a direct approach to the problem, measuring exact participation by users, either at the recreation site itself or through household surveys conducted by mail or by means of personal interviews. Mail Survey Technique Of the various off-site techniques, the mail questionnaire method has much to recommend it. Because it is much less expensive than personal or telephone interviewing and requires no additional staff, the mail survey technique permits broader coverage of the pOpu- 1 lation in question for a given amount of funds. A much larger sample can be taken, which often increases the 1Douglas Crapo and Michael Chubb, Recreation Area Da -Use Investigation Techniques (East Lansing: Depart- ment of Park and Recreation Resources, Michigan State University, January, 1969), pp. 22-23. precisioz addition viewers : response: of the r1 intervie' bias. " thus can they th '1 as Predi herent i it is if! able pre precision of estimates made from the study data. In addition, the bias which can arise from the use of inter- viewers is avoided. Incompetent interviewers can cause responses to be slanted and inaccurate. Also, reluctance of the respondent to respond truthfully in a face-to-face interview presents itself as another potential source of bias. "Respondents have confidence in their anonymity and thus can have a greater sense of freedom to express views they think may be contrary to those held by the majority."1 Mail surveys, however, are by no means infallible as predictors of pOpulation preferences and needs. In- herent in them are sources of bias of potential importance; it is important to examine these, if one is to make reli- able predictions on the basis of mail survey results. The most often asked and significant question with regard to the reliability of mail questionnaires is the matter of non-response. Frequently, conclusions are drawn by researchers on the basis of a small percentage of returns to a mail survey. Are these conclusions valid? The answer to this question lies in a comparison of the required data concerning those who did respond to a particular mail survey with those who did not. If re— spondents and non-respondents differ importantly in their demographic characteristics, interest in the topic under lIbid., p. 23. study , I to repri surveyei has a p governm to dete in a gi is wil this ac such a for the for a C high at study, and the like, then the respondents cannot be said to represent a true picture of the population being surveyed. Non-response in Recreation Research Mail Surveys It is possible that the problem of non-response has a particularly significant application to the work of government agencies attempting to survey their clientele to determine recreation behavior, needs, and desires. It has been suggested that a participant's success in a given recreation activity may have a direct bearing on his willingness to respond to a questionnaire which covers this activity. If so, projections based on responses to such a survey may be exaggerated, because data tends to be for those who participated in the activity the most, not for a cross-section of the population which mirrors both high and low levels of use. Sponsorship of recreation surveys is another factor felt to have a decided bearing on peOple's readiness to respond to them. Some recreation researchers in state agencies feel that surveys under their sponsorship often gain lower response than those under the auspices of a "neutral" university research team. They see fish and game laws as a possible source of antagonism from some recreationists, influencing those subjects not to respond. The question is: Are those who feel antagonistic and disi those wh mental r recreatf of inve: ticular delinea if recr on 11591; importa On the Sults, sponder Charac1 aSpect D0 tho: from t1 prOjEC‘ ana1Y8; mail SI and disinclined to respond significantly different from those who are willing to share information with govern- mental recreation researchers? Problem Statement and Objectives The applicability of social research methods to recreation research, in general, presents a problem worthy of investigation. More specifically, the use of a par- ticular technique, the mail survey, by recreation planners delineates a problem which must be approached and solved, if recreation resource management is to be validly based on user-survey results. The problem is two-fold. First of all, it is important to determine the effect of self-administration on the validity of predictions based on mail survey re- sults. In other words, do data from mail survey re- spondents provide an adequate representation of the characteristics of the population under study? The second aspect of the problem concerns the matter of non-response. Do those who respond to a given survey differ markedly from those who did not respond? The purpose of this project is to approach these two questions through an analysis of data arising from two recreation research mail surveys. between v limited t non-respo made foll views are interviey for reSpc similar t IESponseE did not I I accept tl PErSQnal differ-en Spouses questiOn Obta‘inin other ’ a the trUt havior' Assumptions In the following analysis, a thorough comparison between various response groups is made; the test is not limited to a comparison only of respondent interviews with non-respondent interviews. This thorough comparison was made following the assumption that if non-respondent inter- views are similar to respondent interviews, respondent interviews similar to the mail returns for those chosen for respondent interviews, and those select mail returns similar to the total mail returns, then the total mail responses can be considered representative of those who did not respond at all. Two other assumptions had to be made, in order to accept the results of the study. These are as follows: 1. It is assumed that those peOple selected for personal interviews who were not at home are not markedly different from those who were interviewed and whose re- sponses are used in the present comparison. 2. There is no proof that either the mail questionnaire or the personal interview is capable of obtaining the absolute truth about respondents' activities. The assumption is made that by comparing one with the other, an "averaging” process takes place which approaches the truth regarding an individual's true recreation be- havior, needs, and desires. Studies to Be Examined The problem of non-response can be approached most effectively through an analysis of the actual effect of non-response on a particular study or studies. Two pro- jects undertaken by the Recreation Research and Planning Unit, Department of Park and Recreation Resources, at Michigan State University are well-suited to the necessary investigation. Both studies made use of the mail survey technique, followed by interviewing of a sample of both respondents and non-respondents. With data from these respondent categories available for comparison, the pre- sent study was undertaken to determine, for each project, if respondents answer mail survey questions differently than those asked in a personal interview and whether or not an important difference exists between the respondents and non-respondents to each survey. One of the studies examined is that designed to measure 1968 participation in recreational boating in Michigan, undertaken by the Research Unit as part of a contract with the Waterways Commission, Michigan Depart- ment of Natural Resources. The second project investi- gated is a 1970 study of snowmobile use in Michigan, undertaken by the Unit in cooperation with the Michigan Department of Natural Resources, the United States Forest Service, and the City of Lansing Park and Recreation Department. The each, are d In in each of review repc The purpose findings 0: compare ma Significan Chapter II These studies, with the methods and findings from each, are discussed fully in Chapter III. In order to approach the problem of non-response in each of these studies, it first becomes necessary to review reports of similar research undertaken elsewhere. The purpose of this literature survey is to examine the findings of these studies and the various methods used to compare mail survey respondents and non-respondents. The significant findings of the literature search follow in Chapter II. of rel 0f stu import been f Emerge Projec the dj the v; COmpa: atiOn CHAPTER II REVIEW OF RELEVANT STUDIES In a methodological study such as this, the review of relevant literature has two aspects. First, findings of studies comparing reSpondents and non-respondents are important. Where differences between the two groups have been found, it is important to note which characteristics emerge as having given rise to the variance in most of the projects. If certain variables prove to be the source of the difference in a majority of studies, these then become the variables which should receive close scrutiny in a comparison of respondents and non-respondents to recre- ation research mail surveys. The second topic of concern in reviewing other research on respondents and non-respondents is the actual methods used to compare the two groups. The application of any statistical technique requires that the assumptions of that technique be met. Insofar as these requirements can be satisfied, it is advisable to subject data to tests that are as refined as possible. Thus, differences will be not only measured and reported but weighed for their 10 ll importance. Mail survey non-respondents may differ to some degree from respondents, but the important question is whether the two categories differ enough to invalidate estimates and predictions based on information from the latter. Related below, then, are findings felt to be im- portant because of their consistency among studies com- paring respondents and non-respondents to mail surveys. Following that is a report of the most frequently used methods of comparing the two. Information for the following review has been drawn largely from journals of sociology, psychology, and marketing research. Because survey research techniques have generally not received much attention from recreation researchers, findings of investigators in other disci- plines must be relied upon for guidance in approaching the problem. Findings of Respondent-Nonrespondent Comparisons As pointed out in the introduction to this thesis, recreation research involves some unique factors which may result in a response from peOple who differ markedly from those who do not respond. For this reason, whether or not respondents and non-respondents to an educational survey differ markedly may be no indication of what the outcome might be in a recreation research project. Therefore, strictly relating the number of investigators who found a 12 significant difference, compared to those who found the groups homogeneous, is not apprOpriate here. What is important, however, because of the con- sistency with which it appeared in this review, is the frequency with which particular characteristics were found to differ between respondents and non-respondents in various studies. In studies where parameters for the two groups were shown to differ, the same variables were re- peatedly the source of the greatest variation, regardless of the discipline or subject with which the study was concerned. Because these variables are those which are most likely to exhibit variance in mail surveys, they become the most important basis for comparison in the present study. How Non-respondents Differ Two characteristics which are mentioned most frequently by researchers as being different are education and interest in the survey topic. While many investigators have hypothesized other factors in addition, these two appear quite consistently in all studies where a compari- son has yielded a significant difference. An example is a study done by Suchman and McCandless of Columbia University. They conducted a mail survey in two waves and followed these with a telephone interview of non-respondents. The topic under study was participation in a child-training radio broadcast in Iowa. compari Th1 Whl wi th sc gr in a s 0f boa feld c genera answel and 0( in Wr: t0pic variai Ents 13 When respondent and non-respondent data had been compared, the researchers wrote: The influence of education is quite marked. . . . Whereas almost one out of every two of the respondents with a college education returned the questionnaire the first time, one out of every five with a high- school education and one out of every ten with a grammar school education did so. While educational level might appear more relevant in a study of child develOpment programs than in a survey of boating or snowmobile participation, Franzen and Lazar— feld correlate education with response to mail surveys in general. The two state that, "mail questionnaires are answered more often by people who, due to their educational and occupational background, more easily express themselves in writing, and by people who are more interested in the topic under discussion."2 Interest in the topic of study is the second variable most frequently found to differ between respond- ents and non-respondents. Clausen and Ford, conducting repeated mail surveys of Army veterans, "found a higher response rate from those interested in the survey subject than from other veterans."3 A study of veterans' 1Edward A. Suchman and Boyd McCandless, "Who Answers Questionnaires?" Journal of Applied Psychology, XXIV (1940), 760. 2R. Franzen and P. F. Lazarfeld, "Mail Question- naires as a Research Problem," Journal of Psychology, XX (1945), 294. 3John A. Clausen and Robert N. Ford, "Controlling Bias in Mail Questionnaires," Journal of the American Statistical Association, XLII (December, 19475, 506. educati initial or take zing tI Furthei suppor- higher subjec catior hereir PIEdic tunit; frequ. dicti. 14 educational plans, for example, brought "a much higher initial response from veterans planning to attend school or take training than from those not interested in utili- zing the benefits for which they had applied earlier." Further, studies by these investigators lend additional support to the suggestion that educational levels and higher response rates are positively correlated.1 This degree-of-interest aspect, while measured for subject areas other than recreation, has important appli- cation for the boating and snowmobile studies examined herein. The major purpose of recreation research is to predict future needs and desires for recreation oppor- tunities. If participation is, indeed, reported more frequently by those most interested, estimates and pre- dictions may be too high, so newly built facilities may go unused. Some Findings Suggest No Difference The group of studies represented by the examples above resulted in conclusions that some respondents and non-respondents are markedly different in some respects. It is important to note, however, that this difference does not always exist in a group of survey subjects. For example, Robinson and Agisim, in a market research project on clothes buying habits, refute the 11bid., p. 506. 15 validity of the respondent-nonrespondent dichotomy. In a follow-up inquiry of non-respondents, questions designed to elicit reasons for non-response were included. Only 4.2 per cent of this group gave "Not Interested in Subject" as a reason for not replying. In contrast, 43.2 per cent indicated they had mislaid the questionnaire or simply overlooked answering it. For the most part the non-replying group in this study was made up of people whose reasons for not replying were the result of "physical causes," such as neglect, loss, etc. There are no indications that non-responders were to any significant degree of a different Eype than responders.l The only conclusion one can reach after a review of findings in the respondent-nonrespondent area is that no definitive statement covering all disciplines can be made about the problem. Results vary from discipline to discipline, as constraints on survey subjects vary accord- ing to the topic under study. A generalization that can be made, however, is that where differences between the two groups are found, these differences very often are related to educational level and interest in the topic. Had findings been consistent across the subject areas, investigation of mail questionnaires as a recre- ation research tool would still be necessary, as the work in this particular area is negligible compared to the 1R. A. Robinson and Philip Agisim, "Making Mail Surveys More Reliable," Journal of Marketing, XV (April, 1951), 418. 16 volume of investigations that has been undertaken in psychology, sociology, marketing, and education. Methods Used in Other Studies As the current project is a methodological inquiry, the techniques used to compare respondents and non- respondents are equally as important as the results of these studies. A review of the literature shows the use of several different techniques which vary in their degrees of sophistication. Among these are simple ratings of con- sistency, reporting of percentages, and use of the analysis of variance and critical ratio techniques. The literature search reveals that the method most frequently used involves calculation of the percentages of respondents and non-respondents falling in given response categories for each question, followed by a comparison of these proportions using the chi-square technique. Reid, an educational researcher, conducted a study of sample Ohio schools to measure their use of broadcast equipment. The project was implemented using 3,293 mail questionnaires and intensive telephone and special delivery follow-up of 87 of the study's 1,032 non-respondents. De- fending his sample size, Reid points out, In the polling of non-respondents it is not neces- sary to send questionnaires to all individuals or institutions that failed to answer the original questionnaire. A representative sample can be chosen, and if statistical precautions are 17 observed the responses from this smaller group can be interpreted as representative of the non- respondents.1 Results of this study are useful, in that they show the importance of using techniques more refined than a simple comparison. A glance at some of the percentage comparisons does not reveal much divergence. For example, 71.3 per cent of the respondents indicated their schools made radios available to students, only 3.5 per cent more than the 67.7 per cent of the non-respondents who answered the same question affirmatively. This difference, however, was proven statistically significant at the .05 level, using the chi-square test.2 A methodological study by McDonagh and Rosenblum also employed the chi-square test, but with different results from those reported above.3 The study is im- portant because it goes one step further than other respondent-nonrespondent research up to that time. In addition to comparing mail returns and non-respondent interviews, these sociologists interviewed a sample of 1Seerley Reid, "Respondents and Non-respondents to Mail Questionnaires," Educational Research Bulletin, XXI (April, 1942), 95. 21bid., p. 92. 3Edward C. McDonagh and A. Leon Rosenblum, "A Comparison of Mailed Questionnaires and Subsequent Structured Interviews," Public Opinion Quarterly, XXIX (Spring, 1965), 131-36. mail res1 those th of the 1: questior random 5 who com} For pur‘ view in questio queStic Cthgo] “911 a] liter; at ha: methm 18 mail respondents, comparing their interview answers with those they gave on the mail questionnaire. A systematic random sample composing 20 per cent of the population under study was chosen and sent a questionnaire. When the response period was over, a random subsample of 10 per cent was selected from those who completed the questionnaire and those who did not. For purposes of comparison, key questions in the inter- view instrument were identical to those in the mail questionnaire. Respondents were asked a variety of socio-economic questions and comparisons made between the three response categories. The chi-square statistic in every case was well above the 5 per cent level set by the researchers. There were no significant differences between the responses of the mail questionnaire and those of the interviewed respondents who had not answered the questionnaire. The nonrespondents did not seem to be so selective of some variables as many behavioral scientists assume. The findings of this study imply that researchers should have greater confidence in the questionnaire as an initial tool of research.1 Summary Results of these and other studies reviewed in the literature search have definite implications for the study at hand. The McDonagh and Rosenblum study above is methodologically sound and, therefore, provides a good lIbid., p. 136. model f0] of mail c I may be f: affect S] agencies responde: study, 1: nt's in have not PIOject qUestion the Secu data is the infc Study. designm naire d.- and nOn. methOdO; 19 model for the researcher interested in validating the use of mail questionnaires as a research tool. Because it is strictly methodological, however, it may be free from the influence of certain factors which affect special-subject studies implemented by particular agencies. In other words, while respondents and non- respondents are proven statistically similar in this study, the role of the investigator's image or respond— ent's interest in or commitment to the topic under study have not come into play as much as they might in a specific project undertaken by a government agency. The optimal approach to making a useful comparison of respondents and non-respondents to recreation research questionnaires seems to be: 1. To select an ongoing research project, where the securing of accurate planning and policy formulation data is the major concern, and then 2. To implement the study in such a way that all the information necessary for a thorough investigation of respondent-nonrespondent differences becomes available. This approach is the one adOpted for the present study. Two current recreation research projects were designed and implemented so as to provide mail question- naire data and interview responses from both respondents and non-respondents. This has permitted a thorough methodological examination of a practical survey situation where distin of the catior a IECI 20 where interests, prejudices, and the like could cause distinctly different groups to respond or not respond. Chapter III describes the design and implementation of the two projects whose results were examined for impli- cations concerning the use of the mailed questionnaire as a recreation research tool. CHAPTER III STUDIES SUPPLYING DATA FOR COMPARISON No attempt has been made to compare characteristics of respondents to the 1968 Boating Demand Study with those who responded to the 1970 Snowmobile Study; only different respondent categories within each project have been tested. Still, the two studies make an interesting comparison. The projects were administered quite similarly; many identical questions appear in both. Both made use of interview follow-ups to obtain additional information on respondents and non-respondents. There are differences, however, in sampling procedures and some of the other techniques used, as well as in final response rates. These differences are worth- while to note and suggest possibilities for future research which compares various approaches to the implementation of mail surveys, once the reliability of the technique has been established. 1968 Boating Demand Study The project examined herein is part of an even larger study of recreational boating in Michigan. The 21 Katerwaj Resouro Unit to boating test th of boat three-y by the and th are be i on thes future M Search bESt fC rea$0115 get the lng in 22 Waterways Commission of the Michigan Department of Natural Resources has asked the Recreation Research and Planning Unit to develop a model which will predict the demand for boating opportunities in 1980 and beyond. To identify trends in boating participation and test the accuracy of the model being formulated, studies of boaters and boating participation are being done at three-year intervals. The first project was undertaken by the Waterways Commission in 1966. Using data from that and the present study, factors affecting boating patterns are being "mapped" and examined for their effect. Based on these results, demand for recreational boating in the future will be forecast. 1968 Study_Questionnaire Design In consultation with the Waterways Division, Re- search Unit personnel devised a questionnaire felt to be best for seeking the desired information and eliciting a reasonably high response rate from survey subjects. To get the necessary comprehensive view of recreational boat- ing in Michigan, the following tOpics were covered: 1. Types and sizes of boats and motors used by boaters in the state. 2. Boat storage, transportation, and launching data. 3. Actual use during the 1968 season for different water bodies--inland or Great Lakes. 23 4. Frequency and type of use on the various water bodies. 5. Origin and destination patterns. 6. In-state use by out-of-state boaters and out-of- state use by in-state boaters. 7. Boat ownership and socio-economic characteristics of state boaters.l The questionnaire was accompanied by a cover letter from the Waterways Commission director, making the study a good one with which to examine the question of possible non-response bias introduced when a state resource planning agency conducts a survey of recreationists. Sample Selection All powered watercraft operated in the state must be listed with the Watercraft Registration Division of the Michigan Secretary of State Department. Consisting of 438,017 boaters in 1968, this list served as the popu- lation for the study. In determining a sample size, an analysis of variance in boat-use periods generated by counties of origin (residence) of Michigan boaters was first undertaken, utilizing data obtained from the 1965 study of recre- ational boating in Michigan. Given this information, and the level of response obtained in the 1965 mail survey, it was decided to draw a sample of 21,600 1Ronald Kaiser, "A Study of Multiple Boat Owner- ship in Michigan" (unpublished Master's thesis, Michigan State University, 1970), p. 14. 24 boat owners from the 1968 boat registration records. The sample was stratified by boat length and by county of residence (origin) of boat owners.1 The samples were then drawn randomly within each stratum by the Michigan State University CDC 6500 computer. Study Implementation By the latter part of May, 1969, all the question- naires had been sent out. Because a follow-up check of respondents and non-respondents was anticipated, three control counties were chosen and treated specially. The three selected were Ingham County, for its urban orien- tation, and Grand Traverse and Leelanau counties, for their ample supply of boating Opportunities.2 The questionnaires sent to survey subjects in these three counties used a special technique whereby the identity of the respondent could be determined even if the address on the first page was removed. As mail returns came in, they were matched with a master checklist listing all boaters sampled from the three areas. Following the response cut-off date, six weeks after the final mailing, all those on the master list who had not returned their questionnaires were classified as survey non-respondents. lPaul Fiske, "Boating Demand and RECSYS-SYMAP Simulation Techniques" (paper presented at the Recreation Research Review, Michigan State University, East Lansing, Michigan), p. 4. 2Kaiser, "Multiple Boat Ownership," pp. 23-24. 25 These non-respondents were listed on a new master chart, and a number was assigned to each. Using a table of random numbers, a sample of 200 respondents and non-respondents was drawn, with the in- tention of interviewing 100 members of the combined cate- gories in Ingham County and 100 of the same in Grand Traverse and Leelanau counties together. As the problem of non-response was the major interest in this follow-up, most of the interviewees, 75 per cent, were to come from the non—respondent category; 25 per cent of the follow-up was to be done on respondents to the survey. As illus- trated by the table below, time and budget limitations did not allow for the completion of the desired number of interviews.1 Table 1 summarizes the number of mail returns received from each of the control counties and the number of interviews completed in each. The same questionnaire used in the mail portion of the survey was used for the interviews, so that no bias due to different question ordering or wording would be introduced. The interviews were completed in August, 1969. Data was coded, keypunched, and analyzed cursorily for large percentage differences in characteristics between non-respondents and respondents and between respondents' answers to the mail questionnaire and to the subsequent personal interview. This examination showed only minor lPaul Fiske, personal interview held in February, 1971. 26 TABLE l.--Mail returns and interviews by county, 1968 Boating Demand Study Mail Non-respondent Respondent County Returns Interviews Interviews Ingham 216 34 13 Grand Traverse 64 36 20 Leelanau 35 15 2 Total 315 85 35a aRonald Kaiser, "A Study of Multiple Boat Owner- ship in Michigan" (unpublished Master's thesis, Michigan State University, 1970), p. 24. differences. Still, the overall mail response to the survey was only 29 per cent. Before making boating needs predictions on the basis of this study, it is important to examine these differences more carefully, subject them to statistical analysis, and obtain a sounder basis for con- cluding that the two groups are sufficiently alike to reinforce predictions made on the basis of partial response. 1970 Snowmobile Stugy The 1970 Snowmobile Study has certain things in common with the boating demand study; they differ in some respects, however, and these differences hold implications for future use of mail surveys for recreation research. As the boating study antedated the snowmobile project, designof the latter benefitted from experience with the former. 27 The major difference in the two studies is in their response rates. Whereas only 29 per cent answered the boating study mail questionnaire, 72.3 per cent re- sponded to the questions on snowmobiling. The reason for this difference is that two reminders were sent to late answerers in the latter case. The board spread of per- centage returns to the two studies provides an interesting diversity of conditions under which to test the differ- ences between respondents and non-respondents. If results are the same for both studies, conclusions about the representativeness of partial returns to mail surveys will be strengthened. Snowmobile Questionnaire Design There are many similarities between the snowmobile and boating study questionnaires. The categories for which information was sought in the former are much like those described above, as this study, too, sought to obtain data from which future facility needs and recre- ation preferences might be predicted. The questionnaire covered the following basic areas: 1. Type, ownership, history, and horsepower of snowmobiles in respondent's household. 2. Counties of use and counties of origin for snowmobiling activity during the 1969-1970 season . 28 3. Activities, trips, distances covered, and com- panions on snowmobile trips. 4. Attitudes on snowmobile regulations. 5. Socio-economic data. Samplipg Procedure In April of 1970, there were 128,093 snowmobiles registered with the Michigan Secretary of State. Budget constraints limited the maximum sample size possible to about 5,000, too small to permit a large enough sample in each county so that statistically reliable data would be obtained for every individual county of the state. The procedure finally selected was the following: 1. The state was divided into three regions, from which a total random sample of 5,133 snowmobilers was drawn. 2. Included in this sample were eight counties from each of which a sample of approximately 300 was drawn, hoping to receive responses from at least 200 survey subjects in each county. These large samples were included to get some statistically reliable county- level data. Study Implementation By the end of May, 1970, 5,133 questionnaires and cover letters were mailed out to snowmobile owners across the state. Reminder cards were mailed to late responders 29 on June 16th, and on June 25th, 2,616 questionnaires with a revised explanatory letter were mailed to those who still had not returned their questionnaires. July 14th was designated as the cut-off date for responses, and no further returns were accepted for analysis after that time. A total of 3,705 questionnaires were returned, 3,641 of which proved to be usable. The percentage return rate, 72.3 per cent, was considerably higher than the 29 per cent achieved by the boating study. The follow-up reminders were undoubtedly responsible for this high response rate. After all returns were in, samples of respondents and non-respondents were selected from Ingham and Kent counties, two of the eight counties where 300 question- naires had been sent out. A different form from the mail questionnaire was used in the ensuing telephone inter- views, but the wording of the questions remained the same. In the interview portion of the study, socio-economic data was not requested from those who had responded to the mail questionnaire. Table 2 lists the number of mail returns received in the non-respondent control counties, and the number of telephone interviews completed in each. The data for the mail returns was transferred to data processing punch cards. The interview data was tabulated by hand but not keypunched. 30 TABLE 2.—-Mail returns and interviews by county, 1970 Snowmobile Study No. Reg. Usable Non- County Snowmo- 82mpie Mail respondent §::233223: biles Returns Interviews Ingham 3,448 294 172 48 39 Kent 4,704 310 204 43 35 Total 8,152 604 376 89 74 The data from these two studies is available and lends itself with relatively few problems to a non- respondent-respondent comparison. How meaningful this comparison is depends on the handling of the data and how carefully it is analyzed. Included in Chapter IV is the rationale behind the selection of the characteristics chosen and techniques used in this study for measuring the differences between response categories. CHAPTER IV ANALYSIS OF DIFFERENCES BETWEEN RESPONSE CATEGORIES Characteristics Chosen for Comparison Socio-economic Data The preliminary literature review of other studies comparing respondents and non-respondents revealed that certain socio-economic characteristics often give rise to differences between the two. Of these demographic factors, educational level is the one most often cited as the source of this difference. Therefore, education was felt to be an appr0priate variable upon which to base a comparison of respondents and non-respondents to the Research Unit's boating and snowmobile studies. In addition, total family income of the survey subjects was felt to be an important basis for comparison. If non-respondents possess markedly lower incomes than respondents, they are likely to participate less in recreational activities requiring a purchase of equipment and supplies, such as boating or snowmobiling. If this is the case, it is not valid to predict future participation 31 in the: income Degree Partic: nonresy in boat as the the sur a subje mail qt lated 1 amount aCtivii Studies may Spe If this Ported Pation mobilir mor.e Us actuall may Spe 32 in these activities on the basis of response from higher- income survey subjects. Degree and Patterns of Participation The other parameters chosen for respondent- nonrespondent comparisons were the amount of participation in boating or snowmobiling reported by each group, as well as the geographic distribution of this participation. As stated previously, the degree of interest in the survey topic at hand has often been found to influence a subject's willingness or disinclination to respond to a mail questionnaire. This "degree of interest," when re— lated to a recreation survey, may be reflected in the amount of time people spend engaged in a particular activity. If respondents to the boating or snowmobiling studies are more "interested" than non-respondents, they may spend more time in these activities than the latter. If this is so, projections made on the basis of use re- ported by them will overestimate the amount of partici- pation taking place across the entire boating and snow- mobiling populations. The end result, obviously, is that more use will be predicted for the future than will actually take place and the Department of Natural Resources may spend more money than is necessary to accommodate the state's boaters and snowmobilers. 33 A more detailed comparison of participation by respondents and non-respondents, one which pinpoints the actual geographical location of use by each, is also desirable. For example, respondent data from a lower Michigan county may indicate a high interest in Upper Peninsula snowmobiling, causing a large portion of develop- ment dollars to go into trail construction in upper Michi- gan. If survey non-respgndents from this same area do their snowmobiling closer to home, they may be met with inadequate facilities and maintenance because planners will have assumed that they, too, prefer to go north with their snowmobiles. If the patterns of boating and snowmobiling use by respondents and non-respondents are the same, predictions made on the grounds of information from the former will be accurate and provide a sound basis for planning. Such a similarity cannot be assumed, however; data from both groups must be compared to give a reliable indication that their geographic preferences are the same. Response Categories Chosen for Comparison Both the Boating Demand Study and the Snowmobile Study provide interview data on respondents and non- respondents from certain select counties. Differences or similarities in response groups from these counties will be assumed to represent those throughout the state. 34 Geographical differences in residence, then, will not pre- sent a possible source of respondent-nonrespondent differ- ences in use and socio-economic status. Kent County re- spondents to the snowmobile study will be compared with Kent County non-respondents, and so forth. Still, the process cannot be limited to a simple comparison of interview data from respondents and non- respondents in their particular counties. It is not adequate to assume that respondent interview data accur- ately represents the data collected through the mail, which is the data from which predictions will be made. Two other questions must first be answered: 1. Did respondent interviewees give the same answers in personal interviews as they did on their mail questionnaires? 2. Do those respondents sampled for interviews adequately represent the total number of subjects return- ing their mail questionnaires, or is the data they provide significantly different from that of the total mail returns? For each of the studies, then, where data is available, three sets of comparisons will be made. These are the following: 1. Interview data on respondents chosen for the follow-up will be compared with the information these same subjects gave on their mail questionnaires. These 35 comparisons will be known as "respondent interviews versus respondent interviewee mail returns." 2. Next, data from respondents' personal inter- views will be compared with the data reported by the 5253; number of mail questionnaire respondents in each county. These comparisons will be known as "respondent interviews versus total mail returns." 3. After the adequacy of the respondent interview data has been established, this data will be compared with personal interview data from non-respondents. Format and Use ofguestions 1970 Snowmobile Study Format of Questions.--The snowmobile question- naire was designed to provide information in a variety of areas, two of these being socio-economic characteristics and amount and pattern of use. The particular questions used here for comparison are as shown on the following page. Use onuestions.--These questions, designed for the mail portion of the study, were incorporated into the personal interview, as well. The questionnaire was not identical, however. In the personal interviews, participation was measured for the three most frequently-used counties. .36 . am: no \ N. WQ RN 923 «0 .oz LN?“ ufi g \VV“\ k‘.‘ ”a“: ~AUGQOU uqmzoum no woman mama om: xuaaoo on: once om:.umoe om: now: oumum mo use nonuo HH< ppm «0 zucsoo can no aucsoo umoa mo mucsou A.sap mzo ma xucaou m cw wcwawbosaocm unmam zmv upon no amp sumo assoc “msoz .ocHH Sumo co mama no noses: . 9: E 325 1..mesz .8: use ”SEES ozSHmozzozm «o..— Mmlmm as. um: so» Ea $3228 5:: n . m cowumoso D 8.6 2:. 892» U 3.1.2» - 80.2... D $40.3 - 80.3 D 3.1.3» - 8°wa U 118:2» .- 8063 D «.86., . 08.8 D «8.33 .. 2562 D 23.8. - 80.: D 80.8 $25 Aoco xuoeov 1.33 E 86:38: So» no 828,: $28 BEES “as 33638 Sun ”353358 a; .8 5:5 9 "NH coflumoso So... no 2 3 3 3 2 2 Z 2 a m a o n a DDDDDDDDUDUDDD Aoao sounov Aaaommmaom may no . afim a: 2 893.128 838:8 no mean» .28... as. $9835 .52 33mm mmmzmi m5. mo 5:5 —— .3 coflumoso 37 Out-of-state use was not measured. This presented no real difficulty, since comparative use of Michigan coun- ties for snowmobiling is the interest of the present study. The out-of-state use was subtracted from the mail return data and only in-state participation of respondents and non-respondents was compared. A more serious problem arose from the fact that socio-economic data was not measured in the respondent interviews, if this information had already been obtained from the subject's mail questionnaire. This lack of respondent interview socio-economic reporting prevented a comparison of respondent interviews with the same respondents' mail returns or with the total mail returns. The most workable solution, then, was to compare non- respondent interview information with data from total mail returns, incorporating the assumption that respondent interview data would have been consistent with respond— ents' mail returns and total mail returns. Pattern and degree of use gee covered in the respondent interviews, so these variables could still be subjected to the complete comparison most desirable. 1968 Boating Demand Study Format of Questions.--The boating study, too, sought a variety of information on Michigan boaters. The socio-economic and use questions, somewhat similar to those in the snowmobile study appear on the following page. 38 833 8 8...»; D 83% e 8.8 D 8%» 2 8.8 D .26 B. 838 D 838 2 68.2» D 8.1.8 2 Sod» D 83$ 58: D .33 So 28 0.85. .333 So ESE 2: 025.810 >m 82 mo“. 2282. 5.52.. .25» 50> $52me $5.... 0— ”3 c0383 I III I 29:5 5 o— m— 3 n— N— : o— m a h m m w n N p DDDDDD DDDDDDDDDDDD can So .322 ..~>...: lo 931.. oz» >m Dmhmgmiou 20....(UDom “.0 mm 4<._.O.P ml... mmh538 1'“ 0 w b O N . ~ g E 26 1500 E .38 ZOO .02. .02. .Oz. .02. .Oz. .02. 8 a: :— 25 9.530 ”a.“ 9.35: 8H“. c9:_u_m>iuo..h .Enuhos I8. soon :5 v8: 30> $3 .02 82.3.3 2:30 >420 9.95;: 02....0mzz8 02¢ mugs h «>3 .02 8.3254 Pinon .1: saw... m2MHB UGMHQ $6.5 am.¢a mH.HH mm.ma wm.ma wm.m~ wo.o .ucH .mmmuucoz wm.q wm.- m~.mm wm.ma wa.m mo.h ma.~ Hem: Hmuoa mo.o wmm.ma moa.om ma.mm wa.m~ wo.o we.n Ham: .ucH .mmmm wo.o wo.o~ wo.o~ mo.om wo.oa wo.ca mc.oH .ucH .mmmm EmnmcH u0>o mmw mam can mm mm mm a mmw 0» mam on cam on mm ou mm 0» mm Hops: msouw wmsommmm can hucsoo avoum panama mswumom «mad .mmwuommumo mEooca msowum> cw mmooum mmcommou ucmumwmflv Bonn muomnnam hm>u5m mo wmmucmouomun.m mqmde 63 at a figure suggesting that "30 per cent of all non- respondents have an income below $6,000, a number signifi- cantly different from the rest of the boating p0pulation." The table does give a good breakdown of where survey sub- jects fall with regard to income, but it should not be taken as a statistical measure of differences in response groups. To arrive at a conclusion that is statistically reliable, the categories were collapsed and compared, giving the results tabulated in Table 6. The right-hand column in the table indicates that none of the comparisons are significant at the .05 level. A good example of the value of statistical testing, however, is given by the respondent interview-total mail return comparison for Ingham County. A comparative glance at the percentage of subjects in these two groups falling in each income cate- gory indicates a difference between the two. A statistical comparison, however, proves that the difference is not a significant one and supports the assumption that the difference arose by chance, as a function of sampling variability, and that the respondent data does provide an adequate representation of the characteristics of all those returning their mail questionnaires. 1970 Snowmobile Study As mentioned previously, because no socio-economic data was collected in the respondent interviews, the 64 TABLE 6.--Comparison of total family incomes of various response groups, 1968 Boating Demand Study Count nd % of Low % of High 2c Prob. C m 3:.3 Income Income x of o p lson Subjectsa Subjectsb Sig. Ingham Resp. Int. vs. 60.0% 40.0% Resp. Int. Mail 53.8% 46.2% '037 '7679 Resp. Int. vs. 60.0% 40.0% Total Mail 34.4% 65.6% 2'707 '0999 Resp. Int. vs. 60.0% 40.0% Non-resp. Int. 66.7% 33.3% '142 '7060 Grand Traverse & Leelanau Resp. Int. vs. 50.0% 50.0% Resp. Int. Mail 57.9% 40.1% '203 ‘6526 Resp. Int. vs. 50.0% 50.0% Total Mail 68.3% 31.7% 1°676 '1954 Resp. Int. vs. 50.0% 50.0% Non-resp. Int. 48.3% 51.7% '011 '9156 3 Counties Combined Resp. Int. vs. 54.2% 45.8% Resp. Int. Mail 56.3% 43.8% '024 '8767 Resp. Int. vs. 54.2% 45.8% Total Mail 43.0% 57.0% 1'114 '2913 Resp. Int. vs. 54.2% 45.8% Non-resp. Int. 57.1% 42.9% ‘060 ‘8058 aIncome of under $3,000 to $9,999. bIncome of $10,000 to $25,000 and above. cCalculation indicating significance of test. 65 comparison of incomes was limited to one of non-respondents with the total mail questionnaire returns. Income cate- gories were collapsed to insure that at least five obser- vations would fall in each of the categories, thereby meeting the assumptions for the chi-square test. The analysis, then, compared members of response groups with under $10,000 total family income with those having an annual income of over $10,000, as was done in the boating study. The results are shown in Table 7. Both non-respondent and mail respondent p0pulations have a much greater percentage of high-income than low- income members. The fact that this is reflected con- sistently in both groups, however, makes it possible to accept the hypothesis that there is no significant differ- ence in the total family incomes of respondents and non- respondents to the snowmobile study. Summary Statistical comparisons of data from both the snowmobiling and boating studies reveal no significant differences between the incomes reported by respondents on their mail returns and in their interviews; between the incomes reported by respondents and those of the total number returning questionnaires; and, finally, between the incomes of respondents and non-respondents. 66 TABLE 7.--Comparison of total family incomes of various response groups, 1970 Snowmobile Study County and %Iof Low % of ngh 2 Prgb. Comparison ncome ncome X 9 Subjects Subjects 519. Ingham Non-resp. Int. 17.1% 82.8% vs. Total Mail 22.9% 77.0% '561 '4539 Kent Non-resp. Int. 32.1% 67.9% vs. Total Mail 23.0% 76.9% 1'104 “2934 67 Amount of Recreational Participation 1968 Boating Demand Study To evaluate the differences in boating partici- pation of different response groups, a comparison was made of the mean number of days of Great Lakes and inland boating done by each. The results of these tests are reported below. £1 and xi denote the mean days of participation of the two groups, respectively, being tested. The answer to El - i: measures the actual differ- ence in these means. To test the significance of this difference at the .05 level, the standard deviation of the difference was multiplied by $1.96, covering 95 per cent of the samples. No significant difference was shown if this figure exceeded the actual difference in means. Had the actual difference been larger than this figure, it would have been significant, indicating that survey non-respondents do more or less boating than do respondents. Results of this test are tabulated in Table 8. A look at the first two columns in the table gives a ready indication of how close the amount of partici- pation reported by each group actually is. The next-to- the-last column is an indicator of how much leeway was available before the difference in means would have fallen outside the acceptable range and indicated a 68 TABLE 8.--Comparison of amount of boating of various response groups, 1968 Boating Demand Study i a i- b 2' '3? ii. 9:0 Signifi- l 2 l 2 X -X cant? l 2 Ingham Resp. Int. vs. 27.5 Resp. Int. Mail 19.0 8.55 119.8 no Resp. Int. vs. 27.5 Total Mail 29.8 2.25 112.8 no Resp. Int. vs. 27.5 Non-resp. Int. 23.3 4.24 113.8 no Grand Traverse & Leelanau Resp. Int. vs. 31.8 Resp. Int. Mail 27.4 4.44 115.9 no Resp. Int. vs. 31.8 Total Mail 38.5 6.69 113.9 no Resp. Int. vs. 31.8 Non-resp. Int. 34.9 -3.05 118.3 no 3 Counties Combined Resp. Int. vs. 30.1 Resp. Int. Mail 23.74 7.45 114.6 no Resp. Int. vs. 30.1 Total Mail 32.8 2.69 112.5 no Resp. Int. vs. 30.1 Non-resp. Int. 30.1 -.03 114.8 no aMean boating days of first groups in comparison. b Mean boating days of second group in comparison. 69 significant difference. The actual differences did not approach this level; they are clearly not significant. 1970 Snowmobile Study The same calculations as those described above were used to assess possible differences in snowmobiling participation by those who did and did not respond to this study. The results of these comparisons follow boating study results and are reported in Table 9. The respondent interview-respondent interviewee mail return comparison reveals a consistent discrepancy in the number of days of snowmobiling reported. The interview data shows a higher degree of participation than that reported in the mail questionnaires from the same people. This difference gives some support to the suggestion by some researchers that because of a biasing influence introduced by the presence of an interviewer, subjects will inflate their answers in a personal inter- view. The important factor, however, is whether or not this difference is 13533 enough to make respondent inter-l views a poor representation of all those who replied to a survey. The difference is not significant here, so that the respondent interviews can be considered valid for a comparison with non-respondent interviews. Summary This test is perhaps the most significant one of the study. Many comparisons of reSpondents and 70 TABLE 9.--Comparison of amount of snowmobiling of various response groups, 1970 Snowmobile Study County and Comparison x1 X2 Xl-X2 11.960 SE-x l 2 Signifi- cant? Ingham Resp. Int. vs. Resp. Int. Mail Resp. Int. vs. Total Mail Resp. Int. vs. Non-resp. Int. Kent Resp. Int. vs. Resp. Int. Mail Resp. Int. vs. Total Mail Resp. Int. vs. Non-resp. Int. 38.1 38.1 38.1 27.5 27.5 27.5 29.6 38.0 40.1 19.0 29.8 23.3 8.47 .03 -2.06 2.25 4.24 111.01 110.15 114.32 119.82 112.79 113.88 no no no no no no 71 non-respondents to mail surveys have found the latter's lack of interest in the topic to be a major cause for non-return. If the amount of recreational activity in which subjects participate can be considered directly representative of this degree of interest, a comparison of the activity of respondents and non-respondents shows no difference. For recreation planners allocating development funds, assessing the total statewide amount of partici- pation in an activity becomes the most crucial task. Sample surveys of recreationists are valuable only if their results can be projected to be descriptive of the entire population. For this reason, non-response, if it indicates a lower amount of participation, creates a serious bias. The foregoing test of participation by various response groups shows that, in fact, there is no significant difference in this participation; respondents to the boating and snowmobile studies can be accepted as being representative of all participants in these two activities. More generally, we can accept the hypothesis that there is no significant difference between respond- ents and non-respondents to similar recreation mail surveys in their amount of participation in a given activity. 72 Geographical Distribution of Participation Going one step beyond a measurement of the amount of participation is an analysis of the specific location, county-by-county, of recreationists' activities. The technique most suited to making this evaluation, as stated earlier, is multi-variate analysis. Such a comparison would likely be possible if one were examining respondents and non-respondents from every county in the state. Only one county was under study at a time, however, in the present project. Since most of the use in each case took place in this county, the number of responses in each category was not evenly distributed and did not approximate a normal distribution, thereby failing to meet one of the assumptions of the multi-variate tech- nique. If one were to use this analysis, the low cell frequency would have required so much collapsing of counties that the results would not have been a meaningful indicator of use on an individual county basis. For a sample of this size, therefore, where use is concentrated mostly in one or two counties, the use of descriptive statistics is more appropriate. While not allowing for a statistically significant comparison, this technique does illustrate the exact distribution of use and allows for a reasonable comparison. Employing this method, each questionnaire was coded for the respondent's days of use in each county. Total days of use for one response group were divided 73 into the total use per county by this group, arriving at the percentage of their recreation days which were spent in each county. These figures were transferred to maps of the state, each one comparing two response groups. Maps illustrating the distribution of 1968 boating use in Michigan were not made. Due to the small number of follow-up interviews completed, it is difficult to arrive at an adequate representation of boating use patterns across the state. Maps illustrating the patterns of snowmobile use across the state could be made, as a larger sample of survey respondents and non-respondents was interviewed. This larger sample size provides a more reliable repre- sentation of actual use by the different response groups, making a visual comparison more valuable than it would be in assessing respondent-nonrespondent differences in the boating study. Following the same order of comparison used in the tables, the snowmobile maps are in order by county. For each county there is a series of three maps making three comparisons--of respondent interviews versus respondent interviewee mail returns; respondent inter- views versus total mail returns; and respondent inter- views versus non-respondent interviews. Results of the snowmobile comparisons show a fairly uniform distribution of use by each group and, 74 nu M»! "I A - '1 ounc- l l r- -' . |Immuucm L, g ‘1 L r‘ Inna r": m..c 1‘- - I ' I.‘..”‘"' . Luct . j . I ' _ _ _ _ I L '- - «L - - - .' o ' | . . - - 1 IIOI , uuu r _ 1 I WIMP!" , 5---... ', uuootcnmL—---'a ' ,oucmm' r'__Z_.L. lucxluc I... :3- b I , , nun . I I a H 0% I 00 9 m 'nutrI “'0qu \ 93,! I Y“ S no I :0?“ cunt: onIL - .1 // z 4.0 H I J -_-J'1 Study: 1970 Snowmobile Study County: Ingham at. £41,427: .4 Comparison: Percentage of J’IJHI I 7 total snowmobile days spent 3:21;." Tafiéfifliaiu? in each Michigan county-- I i— 7' ’—'—-4 FL respondent interviews vs. 1':f€J:T§'“' respondent interviewee ° " °°" """““"’°“"°I ------ mail returns r' , -. "-..l - - - «- ounon sum" ' r. k... '0 I r- I -49 Legend: top = Respondent ,.z4, I . r-i interviews -mf L-L_1-- -.—--5""(T*”uwh.cum ovum: ' mm .cuur ”Vanni: I ' | bottom = Respondent I I 2:33}. I ' _ _ _'— --‘ interv . e ' .J L3 0 ' .. -'_ -rouuuo .Iacou‘ 1 wee utnu rum; .7 ("on ' Inn“ :UVM'W". I mail returns 1'4! .39. 3 . 7 . I ...... Lauri-2 3201-44-1? .. --.- m.uu:uuuz.: “tum rucuon T':.;t;u" um: 'I | ' Illlllll l ' I : ' l . I I | I . Figure 1. --Comparison of geographic location of snowmobile use in Michigan counties, respondent interviews versus respondent interviewee mail returns, Ingham County. 75 l L L r’ quu r: «mu: 1. _ I I Inna": ' I-- ---i ' ' L- - .1 MOI , "0 Mull _ _: I o/ 16'0"!“ ' F--— :0, rscumcun'.--- J ' poll-80;: r- - _ _ 4' 1 I neuuc L'J .9 , nun . I ' ‘ QI03 :3: m_-4;4 0Q I“: " Alff-lf- f' 10/780!” Inn“: “Fl". 1:1 . ‘ I o 7 ‘ , /. 3 | Study: 1970 Snowmobile Study rJJ-J----i ------- o §:;j:\an 'IIITI 'cooov § cunt: m L- mm 0860 I - .5. :13"; _/._é’ "°°"‘ County: Ingham . I '-7.-7. I --?— ‘ p.33; .4, 36:66qu amino}; ’ Comparison: Percentage of #31 '4' 7 .._’1’_ total snowmobile days spent -' 9'l'4"° in each Michigan county-- respondent interviews vs. total mail returns 3. 9 l ' : Legend: top = Respondent 239-310 ' .--' - interviews ' :'°'"°‘”' T""'°" ""m 1' - ‘- '1 4 I ' I ._ - ' ./' ’ ' ' -__ruu;u.'-m - bottom - Total mall Sufi-Rfifififimnl. : returns . :.74 , . I W—--L’ ,. ..I- ._ I. - "1/J. .. - -3- -rouuuo .IACOII‘ ALLIIAI E .:.5: :17? g 92,,“ :uvIIIosym. .' Imz 'waIfi3' ' """" rL --- --- r m, “Murat-02.: “no“. ucuon Influx" um: I o ‘ ‘ ' I ' 3;... [.7. “I": ."'c” 'NILL.m' LI"'I' I IOIIOI I : I : I | : . I I | ” "mun ' ' '1 Figure 2.--Comparison of geographic location of snowmobile use in Michigan counties, respondent interviews versus total mail returns, Ingham County. 76 I I I I CHINIIA acumen" L - - - J ., 44¢ Iucxuuc I I L....__.1 P " -I- - - 1 - - - 0W}.- .. 1.1.1575” .wfrg‘m" 4% : ‘3- ' 01 I Study : 1970 Snowmobile Study _ r_ :quII;.;.;,;To:ezo.-h;.:: “I! I]... f."' _ I .3 I I. y . County: Ingham 4:42;, 4-11-“: " I‘d: “I. q'IX'm ”an,“ 30:63.7. &;.:.- n01; .- Comparison: Percentage of 7- :43 If} 571."! :4;- total snowmobile days spent 11 -L4r--- -J-}--L-- in each Michigan county-- respondent interviews vs. non-respondent interviews T - - - -' - - - -' ------- 1 ocuu ,IC'A'OO ucoan II'IOII-Ufnouuo , ’00 '.:—~ ' ' I E.- LeSend: top = Respondent :::J./ L___,, _L_,-I--%3 interviews um: :IOITCALI Ionnon “0qu : h 1" 1.5:; -1' :0 : : '35::5rm‘4- - - '- = - .. - - - — - - - - - -I , bottom Non respondent ”ma: 5”" TcLIIIrou-F“I“‘-. I I." cum lnteerews ,._ I . , . I _ I ' ' I L" ' ' , JILL-93°“; .. _ — - _ —I .. _ ....... '— - ALLIOAI , ”an ' uto- : m...“ :uvmmu' I . .i! .32 4 , . I ...... L- -L’Eé L491: "1. ‘. .m‘.:IALAIA2.1'-¢M_mu IJ.C‘.°. I"."' I." "V.‘ I I I ' 1 . I . I C‘.. I. Int": .u‘.cu .IIILLM' 1""“ .IOIIOI I . I | I l . I I I I l A ___- Figure 3.--Comparison of geographic location of snowmobile use in Michigan counties, respondent interviews versus non-respondent interviews, Ingham County. 77 I uuu r: I CIRCUIT?! I ~ - _ _I we! ' : . ' null r _ _I I Icmrnu I. .. - .. ‘I L I IICIIOOLCIW _ _ - - ,oIcIIII on. r-- __ _ 4 1 O 3 I “cm“: ‘L _. - .- I . 3 .‘ 'IIIT I than E I ,IOI. § CIIAILE ouL -J : -w L-- “In: I' o I I pm» 'mnm . “mu I O ’ I ' Study : 1970 Snowmobile Study 'WEJIEIIJIIQIGJIIFFIEJJ .Tlav. I l . I____ .— County: Kent i'}_' 3.2 4' ' ’ I I r 5,... ‘ ,, Banana-.63.; ‘ Comparison: Percentage of ‘T I : : ~57” I - - - .. - I - -I. - - snowmobile days spent in 1"6 “our“... $5.37,...“ I"—" I I FL I I each Michigan county-- respondent interviews vs. I um respondent interviewee m... ,mmo IJ,§..',[;I'II3¢IJIIEIIII:I‘, I mail returns __ _ ::£ZI . . __--' "4 Léf’.I.- _L_4--!-..-‘-J I Legend : top = Respondent ' 1"”"33; I""'°': ""m I r .3' interviews ,, Imr‘ '} 77 | P -I- _...u,.;; r_’__,'_____r___,ulmu' L--- lunctm ‘ ' IOUII |¢LI“'°' I.III"‘. | . . bottom = Respondent 6.0 : 63.0 I I , I I I 1 o o _ L - - -- - lnteerewee _ r _:_ _ L _ _ _:_ _ _ -E- - "mun I"°°"'I mall returns Atuun I “an ' urou : In...“ :uvnnflu: . ...... L ' 4 ' L _ l _ ._ _I_ I I " T ' ' ' ‘ 1' - ' - " ' 1" VAII IIIIItII,“L“‘l-. cALIIomI . “6"" {nun-u. um , I I | I I I I .ln'!.I---"---'---TJ--I--'-_-I-J--- I can :If-mIM:.n.c. mum'u."u IIIIIIIIIII: I ' L L Figure 4.--Comparison of geographic location of snowmobile use in Michigan counties, respondent interviews versus respondent interviewee mail returns, Kent County. 78 I ILOII 5.-...- ' I '86.;0LCIA'YL-_-_.I ' ' Pcumto?‘ I . ' a I :ucmuc L... "2' . a1 -J--_-l ....... Study: 1970 Snowmobi 1e Study _ _ 'muIA|°u',o.IoscooA I neon flangtuv: I . O I, —— —’ | County: Kent ..-L7§L}é--'JfEJ ........ can "I"? pawns. noscou. to“... I Iooeo Comparison: Percentage of "31'- : _. 9 I‘f‘Z : T"; I-'—- I snowmobile days spent in “,3“: If“; - To‘c‘o'f. 17;; I-ICIHI-I 5;“; - each Michigan county-- ___:égf: t : -: IL respondent interviews vs. o4.If§EJ3;_I;;_J___IuI total mail returns “gyrnuno :lgot‘l’AIIOAIILuImoL‘Io. ‘ ~13 . 7 I I .....—-v I ___. I I -- : = e ndent ___J , 455'- _L_ -!"1J Legent top 1:11:23V18W8 cum: ’24 faunas-Io Tounorf“"“' I r _: - - r.-.:.--I --=—— . ' .. -I- -.-I...;.~ . "g ' '9' '-__IuIIu¢I '- bottom = Total mail 57;; I +r';'fi-T¢:II.T;I.:;M""| I returns .ég:6& ,__—I I I -L -L’..'.-I_ .J_"__ L__ _L _ _ -I-- «run .- nun- Iuny In?“ :IIIoIIAII :u'm": 7; I 0’: ‘ | : | I' I ‘ "‘ “F'ui'lja'” ' ’ T L -' 'I" " ' Rifle"- vaII wan quuz.| cum I «cum: IvuIInIIuI w -— '-"' I ' I |I ltnluI '/ .:/__|___ '___ _I__ -4--- fen-n;- 13 “00"": IuIIcII Ffitwfinuu 1| '°"°‘ I ........ ' _..... I o : o/ : : : Figure 5.--Comparison of geographic location of snowmobile use in Michigan counties, respondent interviews versus total mail returns, Kent County. 79 OUOIITOI fi I uuu :- .. I “not": I -: . i . : I-- ---1 I L- - ‘l IIOI I la 7 Iauu _ _I I ICIIIWI" IIIII:IIIIIIIIIIl r- _ _ _ .I. a I ucIIIIIAc I. _. - _ - , I I I um I 3 I I I r- 4 I" o .I' °0 ° .r‘fi S . In?“ I § cunu oIIIL- ' I .--J1 - - .-' - - .99"; w W... :m.»~~~ I -—-—- I 4‘ J I 0/ I 0 . . ———————————— I— -— — — Study: 1970 Snowmobile Study mfirfl :nuuukn'mnIOlcooA I M... "g‘V-I : I o l . ' -_'_:-' -_-L--- County . Kent ISflg'I/‘giflp‘a‘u mo “30:638-133." I Iosco Comparison: Percentage of I757.I:7 : I:; L total snowmobile days spent I-a- radon I-- “In -I':‘°'T'TI mug- in each Michigan county-- ”5' 'A4 I I r‘ respondent interviews vs. r---J:fz4----5--”" non-respondent interviews m... In»; If?" IIIIIIIIII...I...I _. I z I I - ' I -.I "'14 I Legend: top — Rreliggsizsg "' .. 3'! I‘lzfl?c-A I‘L- I.;GI;I1-OAOIIAI I ’Jl ‘ :rair: «‘0 : : .-:...:.._ bottom = Non-respondent “I I ‘43 r'§;a_+_ "Izfifi‘fl'mufic' I . - ovum . cLIIItou I I interViews o I ”—7? I I I I L — -1”. I I neon I I ' I -I- .. can» I I ————— .I - - - -I... — -— I ALLIOAI run: '5?» In...“ :UVIIIISIUIII I {-3. 6 I I I I I ...... L‘T""‘ L__ _.._.L --—'- VAIIIIIIIIIII “L“‘L' I CALIIOUII T“°"°" fawn-«Tn‘I'u , I I I I IIMIIIL_ __:___:___T;:__ _____ _'._____ I c... IIIMI": .."c” '0" LL L. 'L‘."“ :UDI-JOI .— I I 47 l I L I i :47 I Figure 6.--Comparison of geographic location of snowmobile use in Michigan counties, respondent interviews versus non-respondent interviews, Kent County. 80 more significantly, no more than a 10 per cent difference in use of any county by two different response groups. The larger sample size was undoubtedly a factor. In addition, the snowmobile questionnaire was designed so that use by county was measured in one question and days of specific activities in another. This simpler format resulted in a more nearly complete set of responses with which to make comparisons. The result of this analysis of geographical distribution of respondent-nonrespondent use by county is worthwhile for the percentage comparison it allowed. While the breakdowns cannot be taken as statistically rigid in inferring differences or similarities, they do give an idea that the use patterns of various response groups are fairly similar. This study has made use of much data from two studies and several response groups. The statistical comparisons become quite detailed, but their results are vitally important for those who must justify allocation of dollars and resources for a recreational activity on the basis of partial response to a mail survey of its participants. The findings of the present comparisons are synthesized and their meaning interpreted in the final chapter. CHAPTER VI CONCLUSIONS Problems and Limitations of Study Final results of this respondent-nonrespondent comparison must be interpreted in the light of problems encountered in making the comparison. Findings, especially those that are unexpected, are sometimes better understood if there is a thorough picture of the constraints involved. The most obvious limits placed on the implemen- tation of this project and the use of its findings are those inherent in the use of one researcher's data by another researcher. One of the chief limitations, for example, was the small sample size of boatint study respondents interviewed. Had the respondent-nonrespondent comparison been done immediately, perhaps more funds could have been allocated and a larger sample size gained. Another problem was encountered because of certain slight differences in the data collected on different response groups. The major example of this limitation is illustrated by the data collection on snowmobile survey respondents, where no socio—economic data was 81 82 collected in the follow—up interview. It seems advan- tageous, if possible, to use identical questionnaires in both mail and face-to-face data collection, so that the only possible biases entering can be assumed to have arisen from the method used to implement the question- naire, rather than from differences in question working or ordering which are discrepancies within the instrument itself. Interpretation of Results The results of this study are summarized by the following statements: 1. There is no difference in the educational level of respondents and non-respondents in either the 1968 Boating Demand Study or the 1970 Snowmobile Study. 2. There is no difference in total family incomes of respondents and non-respondents in either the 1968 Boating Demand Study or the 1970 Snowmobile Study. 3. There is no difference in the amount of recreational participation by respondents and non- respondents in either the 1968 Boating Demand Study or the 1970 Snowmobile Study. 4. There appears to be no real difference in the geographical distribution of recreational participation by respondents and non-respondents in the 1970 Snowmobile Study. Respondents and non-respondents to the 1968 Boating Demand Study, particularly those in certain Great 83 Lakes counties, show similarities in their choice of destination for recreational boating although a small sample size prevents the drawing of a final conclusion in this regard. How may recreation planners, especially the Michigan Department of Natural Resources, interpret these results? The overall conclusion which arises is that there is no significant difference in respondents and non-respondents to two recreation studies. Because two different studies and p0pulations were examined with the same results it seems the more general conclusion can be drawn that there is no significant difference between the socio-economic characteristics and use patterns of respondents and non-respondents to recreation research mail surveys of similar populations. The contention has long been that predictions made on the basis of partial returns may not be reliable. Certainly much of the literature in education, psychology, and marketing would indicate that this is true. In another area, that of recreation, these findings in other disciplines may not apply. The present study indicates that only a percentage of response is necessary to draw conclusions about basic socio-economic characteristics and broad use patterns of a p0pulation of recreationists. For predictions regarding these variables, the Boating Demand Study, with a response rate of only 29 per cent 84 appears to be equally as valid as the Snowmobile Study, with its 70 per cent response, though the possibilities for comparison in the boating study were not as full as they might have been. The implication seems to be that study funds, rather than being spread out to cover a wide range of the p0pulation, could be concentrated into the careful design and adequate pre-testing of a survey instrument which is then used on a relatively small portion of the universe under study (the necessary sample size depending on the variance of the data and the amount of detail and break- down required in the analysis). Information on a rela- tively small percentage of survey subjects will be ade- quate, but only if the questionnaire and sampling pro- cedures employed are as systematic and free of bias as possible. It appears this is where funds and efforts should be concentrated, rather than on trying to achieve a high response rate from a large majority of recre- ationists. Implications for Future Studies When making recommendations for future studies of this type, it is necessary to discuss the study on which the comparison will be made. Project leaders anticipating a respondent-nonrespondent evaluation should design their studies to include as large a sample of follow-up 85 interviews as possible. Once again, follow-up question- naire foremat should be identical to that used in the mail survey. Some additional comparisons might be useful, as well. To ascertain the need for specific programming, it would be useful to know if non-respondents use their equipment for different purposes or activities than respondents. The multi-variate analysis suggested for the county comparison in this study would be useful, but it is important to remember that a very large sample will be required to reach the necessary cell frequency, if several activities are under scrutiny. Another possible means of evaluation would be a comparison of the proportion of respondent and non-respondent use accounted for by each individual activity. Another very important comparison would be one of attitudes toward laws and regulations. Respondent support for desired legislation cannot validly assumed to repre- sent the feelings of the entire population, unless non- respondents' attitudes are enumerated and compared with them. More research is necessary before it can be said unequivocably that non-respondents create no source of bias in making predictions about future recreation needs. This study, however, because it examines four variables in two different studies gives very definite support to 86 the premise that recreation planners can justifiably allocate dollars and resources for recreation on the basis of partial response to recreation research mail surveys. SELECTED BIBLIOGRAPHY SELECTED BIBLIOGRAPHY Books Burton, Thomas L., and Noad, P. A. Recreation Research Methods. Birmingham, England: UniVersity of Birmingham, Centre for Urban and Regional Studies, 1968. Champion, Dean J. Basic Statistics for Social Research. Scranton, Pa.: Chandler Publishing Co., 1970. Downie, N. M., and Heath, R. W. Basic Statistical Methods. New York: Harper and Row, 1965. Dunn, 8. Watson. Advertising: Its Role in Modern Market- 1 9. New York: Holt, Rinehart and Winston, Inc., 1969. Handy, Rollo. Methodology of the Behavioral Sciences: Problems and Controversies. Springfield, III.: Charles C Thomas, 1964. Hansen, Morris, H.; Hurwitz, William N.; and Madow, William G. Sample Survey Methods and Theory. New York: John Wiley anHSSons, Inc., 1953. Hoel, Paul G. Elementary Statistics. New York: John Wiley and Sons, Inc., 1967. Jahoda, Marie; Deutsch, Morton; and Cook, Stuart W. 5 Research Methodstin Social Relations. New York: Dryden Press, 1951. Jonsson, Carl Otto. Questionnaires and Interviews. Stockholm: Swedish Council for Personnel Adminis- tration, 1957. Kish, L. Survey Sampling. New York: John Wiley and Sons, 1967. 87 88 Articles Baur, E. Jackson. "Response Bias in a Mail Survey." Public Opinion Quarterly, XI (Winter, 1947-1948), 594-600. Cannell, Charles F., and Fowler, Floyd J. "Comparison of a Self-Enumerative Procedure and a Personal Inter- view: A Validity Study." Public Opinionguarterly, XXVII (Summer, 1963), 251-64. Clausen, John A., and Ford, Robert N. "Controlling Bias in Mail Questionnaires." Journal of the American Statistical Association, XLII (December, 1947), 497-511. Donald, Marjorie N. "Implications of Nonresponse for the Interpretation of Mail Survey Data." Public Opinion Quarterly, XXIV (Spring, 1960), 99-114. Edgerton, H. A.; Britt, S. H.; and Norman, R. D. "Objec- tive Differences Among Various Types of Respond- ents to a Mailed Questionnaire." American Sociological Review, XII (1947), 435-44. Ford, Neil M. "Consistency of Responses in a Mail Survey." Journal of Advertising Research, IX (December, 1969), 31-33. Hansen, Morris H., and Hurwitz, William N. "The Problem of Nonresponse in Sample Surveys." Journal of the American Statistical Association, XLI (December, I946), 517-29. Labovitz, Sanford. "Criteria for Selecting a Signifi- cance Level: A Note on the Sacredness of .05." American Sociologist, III (August, 1968), 220-22. Lehman, Edward C., Jr. "Tests of Significance of Partial Returns to Mail Questionnaires." Rural Sociology, XXVIII (September, 1963), 284-89. McDonagh, Edward C., and Rosenblum, A. Leon. "A Comparison of Mailed Questionnaires and Subsequent Structured Interviews." Public Opinion Quarterly, XXIX (Spring, 1965), 131- 36. Magrabi, Frances M.; King, Marcia P.; and Mead, Meredith J. "Gathering Economic and Demographic Data by Mailed Questionnaire." Quarterly Bulletin of the Michigan Agricultural Experiment Station, Michigan State University, XLIX (August, 1966), 77- 89. 89 Mayer, Charles S., and Pratt, Robert W., Jr. "A Note on Non-response in a Mail Survey." Public Opinion Quarterly, XXX (Winter, 1966-1967), 634-46. Nuckols, Robert C. "Personal Interview Versus Mail Panel Survey." Journal of Marketing Research, I (February, 1964), 11-16. Reid, Seerley. "Respondents and Non-respondents to Mail Questionnaires." Educational Research Bulletin, Reuss, Carl F. "Differences Between Persons Responding and Not Responding to Mail Questionnaires." American Sociological Review, VIII (1943), 433-38. Robinson, R. A., and Agisim, Philip. "Making Mail Surveys More Reliable." Journal of Marketing, XV (April, 1951), 415-24. Shuttleworth, F. K. "Sampling Errors Involved in Incom- plete Returns to Mail Questionnaires." Psycho- logical Bulletin, XXXVII (1940), 437. . "A Study of Questionnaire Technique." Journal of Educational Psychology: XXII (December, 1931), 652-58. Sprowls, R. Clay. "Sample Sizes in Chi-Square Tests for Measuring Advertising Effectiveness." Journal’of Marketing Research, I (February, 1964), 60-64. Stanton, Frank. "Notes on the Validity of Mail Question- naire Returns." Journal of Applied Psychology, XXIII (1939), 95-104. Suchman, Edward A., and McCandless, Boyd. "Who Answers Questionnaires." Journal of Applied Psychology, Vincent, Clark E. "Socio-Economic Status and Familial Variables in Mail Questionnaire Responses." American Journal of Sociology, LXIX (May, 1964), Williams, Allan F., and Wechsler, Henry. "The Mail Survey: Methods to Minimizing Bias Owing to Incomplete Response." Sociology and Social Research (July, 1970), 533-35. 90 Young, R. A.; Holland, I. I.; and Gilmore, A. R. "Getting Better Returns from Mail Questionnaires." Journal of Forestry_(November, 1970), 723-24. Reports Crapo, Douglas, and Chubb, Michael. Recreation Area Day-Use Investigation Techniques. East Lansing: Department of Park and Recreation Resources, Michigan State University, January, 1969. Unpublished Material Cajucom, Edilberto Zalvdea. "Michigan Summer Trail Users: A Pilot Study of User Patterns and Characteristics.‘ Unpublished Ph.D. dissertation, Michigan State Uni- versity, 1971. Fiske, Paul. "An Analysis of Interregional Variation in Recreational Boating in Michigan." Study plan submitted to the Department of Resource Develop- ment, Michigan State University, 1971. . "Boating Demand and RECSYS-SYMAP Simulation Techniques." Paper presented at the Recreation Research Review, Michigan State University, 1969. Kaiser, Ronald. "A Study of Multiple Boat Ownership in Michigan." Unpublished Master's thesis, Michigan State University, 1970. Manderscheid, Lester V. "An Introduction to Statistical Hypothesis Testing." Department of Agricultural Economics, Michigan State University, 1969. Meganck, Richard A. "Recreational Boat Transportation in Michigan: A Study of Use Patterns and Character- istics of Boaters Who Transport Their Boats." Unpublished Master's thesis, Michigan State Uni- versity, 1971. Other Sources Fiske, Paul. External Affairs Coordinator, Institute of Natural and Environmental Resources, University of New Hampshire. Personal Interview, Michigan State University, February, 1971. 91 Lanier, Louis. Associate Professor, Department of Recre- ation Administration, University of Alberta, Canada. Notes on 1970 Snowmobile Study, Michigan State University, 1970. Miracle, Gordon. Professor, Department of Advertising, Michigan State University. Personal interview, February, 1971. Yang, Charles. Professor, Department of Communication, University of Illinois. Class notes from Adver- tising Research, Michigan State University, 1969. Zeigler, Sherilyn K. Assistant Professor, Department of Advertising, Michigan State University. Personal interview, Michigan State University, February, 1971. “Willi!llflllll‘s