TELEPHONE iNTERVlEWING AS A PRIMARY MEANS OF COMMUNICATlON WITH "A CONSUMER PANEL Thesis for the ”Degree of M. A. MICHIGAN STATE UNIVERSITY MARTIN BLOC-K 1.969 _ ‘ TH ssus llllllllllllllllllllllllll 3 1293 LIIIWLIIIWJI LIBRARY Michigan State niversity ABSTRACT TELEPHONE INTERVIEWING AS A PRIMARY MEANS OF COMMUNICATION WITH A CONSUMER PANEL By Martin Block The purpose of this study was to explore the feasibility of using telephone interviews for recruiting consumer panel members and for the subsequent periodic collection of data. Telephone interviewing should provide a cost saving over personal interviewing and a time saving over mail interviewing. This study is intended to be primarily exploratory and descriptive. The hypotheses in the study are: I: No difference exists in panel mortality between a panel using telephone recruitment and a similar panel using personal interview recruitment. II: No difference exists in panel mortality between a panel using telephone diary pick-ups and a similar panel using a mail diary return. III: Operating a consumer panel using all telephone interviewing results in a dollar cost savings over a panel using personal interview recruitment and a mail diary return. IV: No difference exists in diary reporting in a panel using only telephone interv1ewing and a panel using a personal recruitment interview and a mail diary return. Data were collected from two portions of the Michigan State Student Consumer Panel during the Summer of 1968. Two panels, one using only the telephone interviewing (no personal face-to-face contact) and one using ‘ ) an initial personal interview followed by subsequent mail diary returns were compared in terms of mortality, cost and diary reporting. Martin Block The findings of the study indicate that personal interviewing facilitates the initial recruiting of a larger proportion of the original sample. Telephone diary pick-ups, however, reduced the mortality rate after the recruitment interview as compared to the mail diary return. Telephone recruiting was cheaper than personal recruiting, but diary collection by telephone was more expensive than by mail. After five weeks, the cost of the telephone panel nearly surpassed that of the personal-mail panel. The major difference in reporting in the two panels, is the lack of reporting of large items in the telephone panel. Other differences can be explained by demographic and socio-economic differences between the two panels. The conclusion of the study is that, under certain circumstances, the use of the telephone as a primary means of communication in operating a consumer panel may be feasible. TELEPHONE INTERVIEWING AS A PRIMARY MEANS OF COMMUNICATION WITH A CONSUMER BANEL By Martin Block A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Advertising 1969 K; :14. r, 7 9“ /- .1. ‘/- (67 Accepted by the faculty of the Department of Advertising, College of Communication Arts, Michigan State University, in partial fulfillment of the requirements for the Master of Arts degree. Director of Thesis ACKNOWLEDGMENTS I am indebted for the help I received from Professor Gordon B. Miracle in nearly every phase of the panel project and the writing of the thesis. I am grateful to Mr. Keith J. Hess, who began the project and deve10ped the diary and questionnaires; to Mrs. Joyce Butler who typed and edited the final manuscript; and finally to my wife, Rosemary Fox Block who helped with coding, interviewing and patience. TABLE OF CONTENTS LIST OF TABLES O O O O O O O O O 0 O O O O O O O O O O 0 CHAPTER I, INTRODUCTION . . . . Z . . . . . . . . . . . Purpose . . . . . . . . . . . . . . . . . . . . . . Problem Definition . . . . . . . . . . . . . . . . Hypotheses . . . . . . . . . . . . . . . . . . . . Sources of Data . . . . . . . . . . . . . . . . . . CHAPTER II, CONSUMER PURCHASE PANELS . . . . . . . . . . The Uses of Consumer Panels . . . . . . . . . . . . Advantages of Consumer Panels . . . . . . . . . . . Problems in Consumer Panel Operation . . . . . . . The Communication Problem in Panel Operation . . . The Communication Problem in Non—Panel Surveys . . Other Panel Problems . . . . . . . . . . . . . . . CHAPTER III, THE MICHIGAN STATE STUDENT CONSUMER RANEL . General Panel Objectives and History . . . . . . . Panel 1 and Panel 2 . . . . . . . . . . . . . . . . Data Collection Plan . . . . . . . . . . . . . . . The Diary . . . . . . . . . . . . . . . . . . . . . CHAPTER IV, SAMPLE MORTALITY . . . . . . . . . . . . . . Quantitative Mortality: Contact and Response Rates The Panel Sample Versus the Student Population . . Panel Members Versus Panel Losses . . . . . . . . . CIIA PIER v , COSTS o o o o o o o o o o o o o o o o o o o o RecrUitment COSCS o o o o o o o o o o o o o o o o o PiCk'Up COStS o o o o o o o o o o o o o o o o o o 0 CHAPTER VI, PANEL DATA . . . . . . . . . . . . . . . . . Mean Expenditures . . . . . . . . . . . . . . . . . Expenditures Within Product Categories . . . . . . Expenditure Patterns by Day of Week . . . . . . . . Follow-Up Telephone Calls . . . . . . . . . . . . . CHAPTER VII, SUMMARY AND CONCLUSIONS . . . . . . . . . . sumry O O O O O O O O O O O O O O O O O O O O O 0 conCIUSionS O O O O O O O O O O O O O O O O O O O O BIBLIWRAPI‘IY I O O 0 O O O O O O O O O O O O O O O O O 0 APPENDIX 0 O O O O O O O O O O O O O O O O O O O O O O 0 RACE mWHh-r—I 26 26 27 28 3O 32 32 37 41 47 47 A8 52 52 55 S7 61 64 64 66 68 7S TABLE 10 11 12 13 14 15 LIST OF TABLES SUMMARY OF PUBLISHED STATISTICS ON DIARY CONSUMER PANEL MORTALITY I . . . . . . . . . PUBLISHED STATISTICS COMPARING PANEL MEMBERSHIP AND PANEL LOSSES . . . . . . . . DATA COLLECTION PLAN . . . . . . . . . . . . CONTACT AND RESPONSE RATES FOR PANEL 1 (PERSOML'MIL) o o o o o o o o o o o o o 0 CONTACT AND RESPONSE RATES FOR PANEL 2 (TELEPHONE ONLY) . . . . . . . . . . . . . . COMPARISON OF PANEL 1 AND THE POPULATION . . COMPARISON OF PANEL 2 AND THE POPULATION . . COMPARISON OF PANEL MEMBERS PLUS LOSSES AND PANEL MEMBERS 0 I 0 C O O O I O O O O O CHI SQUARE COMPARISON OF THE EFFECT OF PANEL LOSSES ON THE AGE COMPOSITION OF THE PANEL FOR PANEL 1 I O I I O O O O O O O O O O C O CHI SQUARE COMPARISON OF THE EFFECT OF PANEL LOSSES ON THE AGE COMPOSITION OF THE PANEL FOR PANEL 2 O O O O O C O O O O O O O O O 0 COST SUMMARY FOR PANEL 1 AND PANEL 2 . . . . CUMUIATIVE COSTS FOR DIFFERENT DIARY COLLECTION TECHNIQUES . . . . . . . . . . . COMPARISON OF DAILY PROPORTIONS OF FREQUENCY OF PURCHASE FOR PANEL 1 AND PANEL 2 . . . . COMPARISON OF DAILY PROPORTIONS OF AMOUNT OF PURCHASE FOR PANEL 1 AND PANEL 2 . . . . RESULTS OF FOLLOW-UP TELEPHONE CALLS . . . . PAGE 19 21 29 34 35 39 4O 43 45 46 49 51 S9 60 63 CHAPTER 1 INTRODUCTION Purpose The purpose of this study is to explore the feasibility of using telephone interviews for recruiting consumer panel members, and for the subsequent periodic collection of data. The use of telephone interview- ing should involve substantial cost saving over personal interviewing, and time saving over personal and mail interviewing. This study is in- tended to be primarily exploratory and descriptive with the intent of stimulating further research. Problem Definition As Shaffer1 pointed out over 17 years ago, ...under most circum- stances, it would appear that the personal contact is to be preferred to the telephone contact in the recruitment of panel members. This has, however, not been proven empirically and is an area in which some experimental work might well be done, since the use of the phone for just part of the panel recruiting could effect considerable saving." Empirical evidence is still lacking. The use of the telephone in consumer panel operation has been re- stricted primarily to pre-recruitment interviewing, or simply making the arrangements for the personal interview recruitment with some brief ex- planation as to the nature of the project for the benefit of the 1James D. Shaffer. Methodological Bases for the Operation of a Consumer Purchase Panel. Unpublished Ph.D. Dissertation, Michigan State College, 1952. p. 277. 1 2 prospective panel member. Telephone interviewing is also used in emergency problems, in answering questions, or even asking a limited number of addi- tional questions that may have been forgotten or added since the last personal visit. The telephone plays a secondary role to personal inter- viewing, even though telephone interviewing can be used in many situations that the personal interview can be used. The Michigan State Student Consumer Panel, the panel used in this study, was designed to test a number of methodological questions including the use of telephone interviewing as a primary means of communication. The telephone was used for the recruitment interview and the subsequent diary pick-ups, for a portion of the panel. Such use of telephone inter- viewing in consumer panel operation is not evident in the relevant litera- ture. Just as Shaffer stated long ago, telephone interviewing seems to be only considered as a supplementary means of communication along with either personal interviewing or mail interviewing in the primary role. The large consumer panels, such as the panel administered by the Chicago Tribune or the Market Research Corporation of America (MRCA) do not use telephone interviewing as a primary means of communication with panel members. Consumer panels, at least those reported in the literature, generally employ personal interviewing for the initial recruitment. When the panel is geographically widely distributed, a mail recruitment tends to be used. Subsequent interviewing tends to be either personal or mail depending again upon geography, amount of data to be collected, frequency of collection, and other considerations. Telephone interviewing is rarely used in either recruitment interviews or subsequent data collection inter- views. Telephone interviewing should bv much less expensive than personal face-to-face interviews even if nothing more than travel expenses are 3 saved. Telephone interviewing would also avoid the time delay in mail interviewing. Hopefully such savings would in a modest way make the con- sumer panel technique applicable to more problems. That is, problems that would have before been considered too small to warrant the expense of a consumer panel. Hypotheses The primary focus of this study is in answering the question: Is telephone interviewing a feasible alternative as a means of administering recruitment interviews and subsequent data collection interviews in con- sumer panel operation? Consumer panel methodologists probably have in the past relied pri- marily upon the personal interview because of their concern for reducing mortality rates and maximizing the quantity and quality of reporting. The mail interview is used only when the personal interview is not feasible because of cost considerations. The primary reasons, as stated by Shaffer and others for not using telephone interviewing in a primary role in con- sumer panel operation are the fears of poorly established rapport and a lack of face-to-face feedback. Formulating hypotheses in terms of some of these concerns and stating them in their null form: Hypothesis I: There is no difference in panel mortality, or the percent cooperating of the initial sample after N weeks, between a panel using a telephone recruitment interview and a similar panel using a personal face- to-face interview. Hypothesis II: There is no difference in panel mortality between a panel using a telephone interview diary pick-up and a similar panel using a mail return diary pick-up. Hypothesis III: There is no significant cost difference in administering a panel using all telephone interviewing and a similar panel using a per- sonal recruitment interview and a mail diary return. Hypothesis IV: There is no difference in the quantity and quality of diary reporting in a panel using only telephone interviewing, i.e. no personal contact, and a similar panel using a personal recruitment inter- view and a mail diary return. The expectation is that the first null hypothesis will be rejected because personal interviewing should facilitate the recruiting process (because of the advantages of the rapport established in the face-to-face interview). But, the differences between the personal interview and the telephone interview should not be pronounced enough to reject the tele- phone interview as a feasible means of recruitment. The second null hypothesis is likely to be rejected as well because the telephone call placed every time reminding the panel member to turn in his diary might be expected to lessen the mortality effect as compared to the mail return which leaves the remembering entirely up to the respondent. The third null hypothesis is also expected to be rejected because, as already discussed, telephone interviewing should be cheaper than per- sonal interviewing. However, the telephone diary pick-up, i.e. data collection interview, may not be cheaper than a mail return at the cost of a postage stamp per return. The last, and most important, null hypothesis deals with the quantity and quality of diary reporting. It is expected that this hypothesis will be accepted. If so, the telephone interview is to be considered as a S feasible alternative to other methods of collecting diary data. The data gathered by the telephone are expected to be very comparable with any data that might be collected using any other technique. The problem is reliability. The question that needs answering in order to defend telephone interviewing as a feasible alternative is: Does the telephone interview recruitment and diary pick-up provide reliable data from a con- sumer pane 1? Sources of Data Data will be utilized from the two portions of the MSU Student Consumer Panel that were Operated during the Summer of 1968. The samples in this study are both small, and of limited generalizability because they consist entirely of college students. The two panels, of slightly less than fifty members, one using the personal interview recruitment and mail diary return, and the other using all telephone interviewing, both running for five weeks, will be compared. Panel members were recruited through an interview during which a demo- graphic and socio-economic “classification" questionnaire was administered. The reapondents were then asked to keep a diary of their purchases, which then was either "delivered" over the telephone to the interviewer or dropped in the mail. The two panels will be compared for (l) gpantitative mortality at three different stages, (2) qgalitative mortality in terms of the differ- ences between panel members and the MSU student population based upon university records, and the differences between panel members and panel losses, (3) the variable costs of administration of each panel, and (4) gpantitative and qualitative diary reporting, in terms of mean expendi- tures per day, expenditures within specific product categories, and weekly purchase patterns. CHAPTER II CONSUMER PURCHASE PANELS After a search of the relevant literature, sources of two general _types were found, (1) those that deal with non-panel communication problems, i.e. interviewing techniques, and (2) those that deal with panel Operation problems other than the communications problem. With the exception of the dissertation of James D. Shafferz, the literature does not contain substantial discussion of alternatives to personal and mail interviews with respect to consumer panel operation. The Uses of Consumer Panels A consumer panel is a tool for the collection of a wide variety of market data. A consumer panel is a sample of "consumers", i.e., individuals or organizations who use products and services, who are in- terviewed more than once and usually on a regular basis. The example of a consumer panel discussed in this study is a consumer purchase panel which involves frequent (daily or weekly), regular, and continuous report- ing from the consumer sample of their purchasing behavior. There are three common usages of consumer panels in collecting market data: 1. Audience Measurement. The number and types of individuals who view TV programs, listen to radio programs and read newspapers and mag- azines, constitute valuable information for the marketer and advertiser. 2Shaffer. Methodological Bases. 3Harper W. Boyd, Jr., and Ralph L. Westfall. An Evaluation of Continuous Consumer Panels as a Source of Marketing Information. .American Marketing Association, 1960. pp. 7-11. 6 Some examples of this type of consumer panel are the A. C. Nielson Radio—Television Index which measures the size and nature of the tele- vision and radio audience using a mechanical recording device attached to the TV set or radio of the panel member, and the American Research Bureau's (ARB) TV Panel which records TV viewing by means of a self-ad- ministered diary. 2. Product Testing. Consumer panels, usually set up as a product usage test, can determine the reaction of consumers to a new product or to a modification of an old product. This particular use of the con- sumer panel technique is similar to a before-after experimental design. The panel is recruited, an initial measurement taken, the new product introduced, and another measurement is taken. This procedure can be continued forming a chain, using one panel to test many products. 3. Purchase Behavior Measurement. Measuring purchase behavior is considered by many (e.g. Boyd and Westfall) as the greatest single asset of the consumer panel technique. The continuous recording of consumer expenditures can provide valuable insights into consumer ex- penditure patterns through time. Such recording can be used to draw market profiles, i.e., the number and kinds of households (consumers) who comprise a market, which kinds of consumers are light users or heavy users of a given product, which families have tried a given product, discontinued the product's use or have become loyal regular buyers. Consumer panels can provide insights into major consumer shifts, such as a move to the suburbs, or the effect of the first child in family on food purchasing behavior, as two examples. Consumer purchase panels can indicate through time; (1) size of the market, (2) composition of the market, (3) brand or store share of market, and (4) seasonality or other patterns. 8 Advantages of Consumer Panels Most of the advantages of the consumer panel technique are con- tingent upon the fact that the panel design is dynamic, a through- time or longitudinal measure of change and trends; rather than a one- time static, or cross-sectional survey. Some of these advantages are: l. Changers and Non-Changers can be readily identified. Over a period of time those individuals who change with respect to some variable can be isolated from those who do not, and the data can be analyzed or compared individually. 2. A succession of variables can be introduced into the panel, or portions of the panel as described above in the section on product testing. The entire panel can be divided into experimental groups, and control groups, employing the before-after with control group experi- mental design. 3. Panel members, because they are familiar with the system after being a panel member for a period of time, may be more willing to submit to "exacting" interviews than they would be for a one-time interviewer, who probably would be unfamiliar in both person and technique to the reSpondent. Generally panels can collect more data from a given respond- ent than can other survey techniques. 4. If self-enumerative techniques, such as a diary, are employed, the data tend to be more accurate because the respondent relies less upon his memory in recording the data. 5. Panels have a cost advantage over repeated personal interviews, especially if the data are desired at frequent intervals. This applies only to those situations of comparable size and budget. 9 Problems in Consumer Panel Operation As a guide to consumer panel operating technique, Shaffer and Quackenbush have prOposed four problem areas in consumer panel Opera- 4 (l) the sampling problem, (2) the communication problem, (3) tion: the reporting problem, and (4) the tabulating problem. The focus of this study is on the communication problem. The Communication Problem in Panel Operation The communication problem in consumer panel Operation as described by Shaffer5 consists of five phases: (1) establishing contact with potential respondents, (2) establishing sufficient initial cooperation and rapport to get respondents to listen with interest, (3) asking the questions in a way that are understood by respondents, (4) maintaining sufficient cooperation and rapport so as to stimulate accurate response, and (5) receiving and accurately recording the responses. These phases can be shortened to two essential tasks, (1) the re- cruitment effort and getting initial cooperation, and (2) maintaining cooperation and collecting or picking-up the data. Each of these tasks directly involve communicating with respondents. As already mentioned, there are three ways of communicating with reapondents: (1) the personal face-to-face interview, (2) the telephone interview, and (3) by mail. Shaffer discusses personal interviewing in terms of fulfilling both tasks. Personal interviewing is most widely used during the first task, or recruitment effort. Personal contact is generally thought to be the James D. Shaffer and C. G. Quackenbush. ”Cooperation and Sampling in Four Years of M.S.U. Consumer Panel Operation" Quarterly Bullegip. 38 (August 1955) p. 85. SShaffer. Methodological Bases. pp. 229-230. 10 preferred method, and is used wherever possible. The major exception being the use of mail in large national panels. Maintaining OOOperation can also be accomplished through the personal 6 Periodic visits to check on the respond- interview, according to Shaffer. ent, ”to see how they are doing," has been used most effectively. Personal contact can also be used as a means of picking up data by having the inter— viewer make a visit at the end of each reporting period. Personal pick-ups run into the problem Of finding the panel member frequently not at home, causing expensive call-backs, but provides the advantage of maintaining a greater degree of cooperation.7 The decision to use the telephone method of interviewing, Shaffer claimss, is essentially economic. Shaffer goes on to say that telephone interviewing can generally be used wherever personal contact can be used, and is normally much less expensive. Shaffer9 describes three important disadvantages or problems of recruiting panel members using the telephone method: (1) more difficult to hold the respondent's attention, (2) more difficult to establish rapport, and (3) interviewers lack the advantages of interpersonal face-to-face feedback in handling the interview. The main advantage of telephone interview recruitment is that the telephone facilitates contact with some families that are difficult to reach through a personal interview. Although, as Shaffer points out, the preference for the personal interview is clear, this preference is not founded in data. Telephone contacts are commonly used in the pre-recruitment phase, not as the sole interviewing method, but to facilitate later personal 61bid. p. 246. 71bid. p. 250. 81bid. p. 274. 91bid. p. 276. ll contact.lo The telephone also finds frequent use in the post-recruit- ment or maintenance phase.11 The telephone can be used for; (1) encourag- ing regular reporting, (2) for acquiring additional information, (3) for communicating instructions and answering questions, and (4) for building and maintaining rapport. Typically telephone interviewing is not used to pick up data, and is generally regarded as'a facilitating tool rather than a primary interviewing technique. 12, is used primarily in Interviewing by mail, according to Shaffer the pre-recruitment and post-recruitment phases. The introductory letter is a common example of mail used in the pre-recruitment phase. Letters of welcome, newsletters, answers to personal inquiries, are all fairly common examples of the use of mail in the post-recruitment phase.13 Mail is used extensively in recruiting in large national samples.14 Mail offers the advantage of being able to cover a wide geographic area, which is the primary distinction of a national panel. Mail however pre- sents the problem of low return rates, which greatly raise the cost of each panel member recruited. Normally, a mail recruitment effort includes an elaborate mail pre- and post-recruitment plan and some material in- centive plan. The Communication Problem in Non-Panel Surveys As is the case in panel surveys, the personal interview is held as the standard method of interviewing in most of the literature dealing 10Ibid. p. 282. 111616. pp. 283-286. 12Ibid. p. 289. 13Ibid. p. 306. 141618. p. 308. 12 with non-panel surveys. The other techniques, telephone and mail, are generally discussed in terms of their comparability to the personal interview. Baeza15 lists four advantages of the personal interview: (1) it allows better control of the sample, (2) more questions can be asked, (3) field work can be done fairly quickly, and (4) much more informa- tion can usually be obtained by the use of the personal interview. The disadvantages Baeza16 lists are: (1) personal bias introduced by the interviewer, and (2) interviews tend to be given too hastily. Looking for alternatives to the personal interviewing technique, Hochstiml7 conducted a survey using all three techniques and found that the return rate was high for all three methods, questionnaire completeness was high for all, costs were highest for the personal interview, with telephone and mail significantly lower in cost with telephone being only slightly more costly than mail, and most important, very little difference in terms of validity between the three methods. Cahalan18 found high response rates, but found the refusal rate for a telephone interview was more than twice that of a similiar personal interview. 15Marco A. Baeza. Sampling and Response Differences For Three Methods of Enumeration Obtained in a Study of Consumer Potato Preferences. Unpublished Ph.D. Dissertation, Michigan State College, 1950. p. 3. 16Ibid. p. 3. 17Joseph R. Hochstim. "Alternatives to Personal Interviewing" Public Opinion Quarterly. 27 (Winter 1963) 629-630. 18Don Cahalan. "Measuring Newspaper Readership by Telephone: Two Comparisons with Face-to-Face Interviews" Journal of Advertising Research. 1 (December 1960) 1-8. 13 Despite Hochstim's Optimism, telephone interviewing seems to encounter a problem of higher refusal rates than personal interviewing. Schmiedeskamp19 says about his experience with telephone interviewing, that in many instances the personal and telephone data cannot be dis- tinguished from each other, and that in many cases there is no evidence that the more than 20% non-response in the telephone portion made a significant difference in the results. Perry20 found that using a telephone directory for a sampling frame allows for the possibility of arriving at a precise estimate of the error involved. He suggests that, while the use of telephone directories has been assailed among social scientists, it may be worthwhile to consider the possibility of using the telephone directory. Sampling may not be any more of a problem for the telephone survey than it is for the personal interview survey. Shaffer21 lists the major advantages of the telephone interview in cross-sectional surveys as follows: (1) low per unit cost, (2) ease of getting interviews with upper income groups, (3) the rapidity with which the data can be obtained, and (4) the simplicity Of the procedure. Baeza22 adds that the telephone interviewing technique also offers the advantage of spreading interviews within an individual city in random fashion. 19Jay W. Schmiedeskamp. "Reinterviews by Telephone" Journal of Marketing. 26 (January 1962) p. 32. 20Joseph B. Perry, Jr. "A Note on the Use of Telephone Directories. as a Sample Source" Public Opinion Quarterly. 32 (Winter 1968-1969) p. 695. 21Shaffer. Methodological Bases. p. 275. 22Baeza. p. 3. l4 Shaffer lists the disadvantages of the telephone interview as 23 (1) no truly representative cross-section of the public follows: can be obtained, (2) any decision about how to classify "not-at-homes" will be misleading, (3) there is no adequate way to handle "busy" signals, (4) it is frequently difficult to hear and understand telephone con- versations, and (5) the questionnaire must be short. Baeza24 adds two more disadvantages, (1) only a restricted amount and type of information can be secured, and (2) the difficulty experienced in trying to determine the age, economic condition and occupation of the respondent. Shaffer25 in discussing the relative merits of a personal interview versus a mail interview, says that the personal interview is better, and in a study conducted by him found personal interview results to be statis- tically significant from mail interview results. The question, according to Shaffer, is: Is the personal interview better enough to warrant the additional cost? Nuck01326, in a comparison of a personal interview and a mail panel, found that the reporting of total family income tended to be higher, as well as the reporting levels of education, in the mail panel. There were not significant differences found between the reporting of age between the mail panel, the personal interview, or the census estimate for the two samples. Mail surveys tend to exhibit an education and income bias, that is people with higher education levels and incomes seem more likely to respond. 23Shaffer. p. 275 2“Baeza. p. 3 . 25James D. Shaffer. "Estimating Population Characteristics by Mail Survey" Journal Of Farm Economics. 41 (November 1959) 833-837. 26Robert C. Nuckols. "Personal Interview Versus Mail Panel Survey" Journal of Marketinngesearch. 1 (February 1964) 11-16. 15 Low return rates are one of the most serious problems with the mail technique. Baeza27 reports return rates of 97% for the personal interview, 78% for the telephone interview, and only 4.2% for the mail questionnaire. This low return rate for mail questionnaires, according to Larson and Catton28 , contributes to a mail-back bias. Larson and Catton also point out that the difference between early and late returns are indicators of differences between returns and non-returns. Suchman and McCandless29 report that: (1) as interest and familiar- ity with the tOpic under investigation increase, so does the return rate, and (2) the greater the education of the respondent, the greater the return rate. Education and interest seem to be very important factors in considering the use of a mail interviewing technique. Shaffer30 lists the following advantages for the use of the mail technique in cross-sectional surveys: (1) low per unit cost, (2) a wide geographic distribution of respondents possible, (3) useful in reaching specific classes of peOple, (4) there is no interviewer bias, and (5) no identification of respondents is necessary making it possible in some cases to obtain more honest replies. Benson31 adds that mail interview- ing; (1) may provide more reliable answers in certain subject areas, and (2) if the opinions of all members of a household are required, this can 27Baeza. p. 70. 28Richard F. Larson and William R. Catton, Jr. "Can the Mail-Back Bias Contribute to a Study's Validity?" American Sociological Review. 24 (April 1959) p. 245. 29Edward A. Suchman and Boyd McCandless. "Who Answers Question- naires?" Journal of Applied Psychology. 24 (December 1940) 758-769. 3oShaffer. Methodological Bases. pp. 289-290. 31Lawrence E. Benson. "Mail Surveys Can Be Valuable" Public Opinion Quarterly. 10 (Summer 1946) pp. 234-235. 16 be accomplished more easily through a mail questionnaire. Baeza32 claims that respondents take more care in providing information in a mail survey. The disadvantages of a mail survey as described by Shaffer33 are: (1) it is difficult to obtain a really representative list of the universe required, (2) those who reply are probably not typical members of the list, (3) the questionnaire must be short, (4) it is difficult to obtain detailed qualitative answers or know precisely the offered verbal responses meanings, and (5) it is impossible to know whether the intended person answered and whether or not he consulted others. Benson34 adds that (1) less educated people fail to return questionnaires, (2) a proportionately greater number of replies may come from those biased in one direction, i.e., those who may hold extreme views or strong Opinions relative to some tOpic, (3) informational questions attempting to determine areas of ignorance cannot be used, (4) the value of question sequence is destroyed because the respondent may review the questionnaire, and (5) it is difficult to get full and complete reactions or opinions, and questions 35 mentions concerning vital statistics are usually quite limited. Baeza two other disadvantages; (1) there is usually a high cost per return, and (2) a mail survey requires a longer time to complete. 32Baeza. p. 3. 33Shaffer. Methodological Bases. p. 290. 34Benson. p. 235. 35Baeza. p. 4. 17 Given the advantages and disadvantages of each of the three survey methods, Baeza36 concludes that it is possible to obtain comparable in- formation by the three methods. Other Panel Problems Bucklin and Carmen37 mention three kinds of panel problems; mortality bias, conditioning bias, and reporting bias. These three problems, plus a fourth dealing with the cost problem, comprise the major concerns of the panel methodology literature. 1. Qggg. Most consumer panels tend to be large, high-cost Operations, that require large samples, large well-trained staffs, and large budgets. Consumer panels generally make extensive use of personal interviewing, either material or monetary incentives, and SOphisticated data processing techniques. All of these things make the consumer panel technique im- possible for the low budget researcher. To quote Bucklin and Carmen, "undertaking marketing research by means of a panel is a highly expensive "38 Cost is a very serious problem when cOnsidering using prOposition. the panel technique. 2. Mortality Bias. Deviations from the originally drawn sample, whether random or not, can occur in three places or "stages" when dealing with a consumer panel; (1) selected individuals refuse to COOperate at the time of recruitment, (2) shortly after recruitment, or (3) sometime later in the panel. The problem of mortality bias is then much more 6 3 Ibid. p. 72. 37Louis P. Bucklin and James M. Carman. The Design of Consumer Research Panels: Conception and Administration of the Berkeley Food Panel. Berkeley, 1967. pp. l4-20. 38Ibid. p. 66. l8 ,severe when using a panel technique than when using a cross-sectional survey, simply because the respondent has more opportunities to "drop” out of the study. 39 explain, panel mortality can occur at any As Bucklin and Carmen of the three stages already mentioned. First-stage mortality includes those individuals who refuse at the time of recruitment. Also included in first-stage mortality would be potential reSpondents who cannot be reached to be interviewed, i.e. non-contacts, and those who for some reason are considered as being ineligible for the study. Second-stage mortality includes those individuals who agree to participate, but fail to do so in the first few weeks, usually around three, or the first "month". Third-stage mortality, or steady-state mortality, occurs after the panel has stabilized and usually can be attributed to "natural causes" such as death, hospitalization and relocation. Panel members who make it into the third-stage are generally interested enough in the study that they will not drop out for reasons Of lack of interest, or forgetting, or not enough time, which are the typical excuses given for people leaving in the second stage. In Table l, a summary of most of the published data on diary, i.e. self—enumerative, consumer purchase panel mortality, indicates for example that first-stage mortality averages about 40%. Second-stage mortality typically runs about 15%, leaving a stable panel of between 40 and 50% of the original sample. Bucklin and Carman40 point out that third stage mortality runs about 1 to 2% per month of the original sample. 391bid. pp. 14-16. 4OIbid. p. 15. 19 Dropouts during first three weeks 2nd. Stage 10 10 10 15 26 12 l7 l6 17 Reporting after third week 3rd. Stage 49 48 53 10 15 59 4l 63 67 61 57 30 Table 1 SUMMARY OF PUBLISHED STATISTICS ON DIARY CONSUMER PANEL MORTALITY Name of Panel, Author, Refusals Year, (Source Number). lst. Stage MRCA, New York, MRCA, 1953, (59). 43 MRCA, Early National, Stoneborough, 1942, 42 (83) MRCA, Mature, Sudman, 1959, (86) 37 Pittsburgh, Lewis, 1948, (55) 80 Harvard, Allison, Zwick and Brinser, 1958, 70 (2) Michigan SRC I, Sobol, 1959, (81) 15 Michigan State, Quackenbush and Shaffer, 1955, 47 (80) Decatur, Illinois, Ferber, 1953, (28) 20 New York Sun, Black, 1948, (10) 17 Chicago Savings, Ferber, 1959, (27) 22 University of Illinois, Com. Student Econ's, 1961, 35 (84) Berkeley Food, Bucklin and Carmen, 1967, (12) 64 Average 41 *Table taken primarily from Bucklin and Carman. 13 p. 46 20 The high mortality rates found in the first two stages are explained by at least three factors. Lack of interest, according to Bucklin and CarmanAI, is one of the major predictors of mortality. The respondent who does not get interested in the project, and does not keep the diary regularly, generally drops in the second stage. Other studies, such as 142, confirm the importance Of getting the respon- the one reported by Sobo dent interested in the topic under study. Shaffer refers to this process as rapport building. Another predictor is age. Bucklin and Carman point out that old-age seems to be a predictor of mortality. In Table 2, the ages of panel members are presented from the study reported by Sobol, which indicates that individuals at both extremes, the very young or under 25, and the old or over 65, tend to drop out of the panel more frequently than do people with ages in between. Over an extended period of time, this could introduce a serious age bias. Low education is another predictor of non-response. The illiterate and poorly educated are traditionally among the most difficult to reach for the survey researcher. If income and occupation can be interpreted as some index of education level, Sobol found that the unskilled laborer tended to drOp out of the panel much faster than the professional person, as is evident in Table 2. The higher the income level, the greater the chance of the respondent remaining in the panel, again according to Table 2. If this is the case, then consumer purchase panels tend to concen- trate on the more educated, and higher income families. This bias is “1 Ibid. p. 17. 42 Marion Gross Sobol. "Panel Mortality and Panel Bias," American Statistical Association Journal. 54 (March 1959) 52-68. 21 Table 2 PUBLISHED STATISTICS COMPARING PANEL MEMBERSHIP AND PANEL LOSSES OCCUPATION (1) Panel Panel Difference Members Losses Professional 10% 7% +3% Self-Employed 15 14 +1 Clerical & Sales 12 ll -+l Skilled 36 34 *2 Unskilled 12 16 -4 Retired 6 7 -1 Housewives 6 6 -1 Others 1 1 Not Ascertained 1 1 AGE (2) Under 25 4 7 -3 25-34 20 19 +1 34-44 25 21 +4 45-54 22 19 +3 55-64 l7 17 65 and over 10 15 -5 Not Ascertained 2 2 INCOME (3) Under $2,000 10 14 -4 2,000 - 4,999 44 43 ' -1 5,000 - 7,499 27 22 +5 7,500 and over 16 12 +4 Not Ascertained 3 9 -6 *Table from Marion Gross Sobol, ”Panel Mortality and Panel Bias," American Statistical Association Journal. 54 (March 1959), (l) p. 59, (2) p. 57, (3) p. 60. (Percentages do not add to 100 because of rounding error) 22 critical for the consumer purchase panel because of the link between income and expenditures or purchases. All Of the authors already discussed - Shaffer, Bucklin and Carman, 43 - emphasize the importance of incentive plans and Sobol and Ferber other rapport building techniques as a partial solution to this problem. Incentives are generally considered to be most helpful in curtailing some of the panel mortality loss. Stoneborough44 in analyzing the reasons why people drop out of a panel found that they can be classified into four categories: (1) sub- jective factors, rare in later months or third stage, usually caused by misjudging of the nature of the wOrk, i.e. diary keeping, (2) factors operating in the home, lack spouse permission or difficulty in getting family cooperation, (3) external factors, such as the natural causes mentioned earlier usually found in later months or the third stage, and (4) the functionally illiterate, who are unable to submit usable reports. The kinds of individuals who become panel members, according to Glockas, and who participate continuously have different characteristics from those who fail to become panel members or, becoming a member, fail to participate continuously. In addition to the differences already demonstrated by Sobol, Glock points out the group who participates in the panel continuously exhibit different attitudes, opinions and behavior towards and in reSpect to the panel topic, than the group who fails to 43Robert Ferber. Collecting Financial Data By Consumer Panel Techniques. Urbana, 1959. 44Thomas H. W. Stoneborough. ”Fixed Panels in Consumer Research," Journal of Marketing. 7 (October 1942) p. 135. 45Charles Young Glock. "Participation Bias and Re-Interview Effect in Panel Studies" Dissertation Abstracts. 12 (1952) 756. 23 participate continuously. Glock also claims that these differences are not accounted for by variations in the sex, age, education, and family size Of the groups. The mortality problem is most complex. The precise nature of the two groups, participating and non-participating, in a consumer panel can have a profound effect on the generalizability of the findings. 3. Conditioninngias. A person who originally may not have a clearly defined attitude on the tOpic under study, may receive motiva- tion, under repeated interviewing, to formulate a clear position that he otherwise might not. This most commonly manifests itself as the panel member viewing himself as an ”expert" and acting the way he thinks experts should act, rather than the way he normally would act himself. This problem, the conditioning effect, is not nearly as crucial as is the mortality problem. As Sudman46 points out, while conditioning effects may at times be detected by delicate analysis of variance tests, they have proved to be of no real importance in affecting diary panel reports. Although it is possible to induce conditioning by deliberately calling attention to a specific product, this effect usually wears Off in one or two months. Conditioning, in the area of actual reporting is not a very important problem. Glock47 calling the conditioning problem the re-interview effect, finds no evidence of a re-interview effect on level Of interest in amount of exposure to information about, or level of knowledge of the panel topic. Glock points out that his results show no evidence of an effect of re-inter- viewing on the consistency of response to the same question asked repeatedly 46Seymour Sudman. "Maintaining a Consumer Panel" Proceedings of the American MarketingAssociatiopi 42nd National Conference. June 1959. p.326. 47Glock. p. 756. 24 from one interview to the next. Glock does mention that re-interviewing influences the panel to resolve controversial issues and to do so more quickly than the population it is intended to sample. Bucklin and Carman48 observe that the interest in the conditioning bias problem is great, but the empirical evidence is not. The theory of conditioning bias, as is pointed out, is clear enough, but according to the data the effect of conditioning, if any, is slight indeed. 4. Reporting Bias. SlOppy responses, consistently-biased, randomly-biased reaponses; as well as recalling and recording behavior may contribute to reporting bias. Sloppy responses usually occur when the respondent loses interest in the project and fails to report accurate- ly, or write in a legible hand. Consistently-biased responses may be caused by the form of a question, or the method of administration, or a constant bias located in the personality of the respondent. Memory decay also plays a very important role. The reporting problem is more difficult than the conditioning problem. According to Bucklin and Carman49 the topic or reporting has been adequate- ly dealt with in the literature, and is not necessary to report all of the findings here. Important here is the differences in quantity and quality of reporting, or "completeness" of diary reports. Lewis50 found that the reporting is significantly better if the reporting period is weekly 48Bucklin and Carman. p. 18. 491mm. p. 20. soHarrie F. Lewis. "A Comparison of Consumer Responses to Weekly and Monthly Purchase Panels" Journal of Marketing. 12 (April 1948) 449-454. 25 51 confirmed these rather than monthly. Allison, Zwick and Brinser findings. Other manipulations in panel methodology, other than varia- tions in the questionnaire and diary forms themselves, have not been dealt with to any great extent. Completeness of diary reports, for example, have been related to the interest of the respondent by Bucklin and Carmansz. Sloppy re- porting has been related to lack of interest in much the same way as lack of interest in the panel topic has been related to panel mortal- ity. The effect of rapport-building techniques on sustaining interest in the panel, thus maintaining a relatively high degree of completeness in the reporting in the diary forms is implied in the literature, but not as clearly shown in terms of data. The relationship of the communication problem in consumer panel operation to the reporting problem for the most part is clearly "theoretical," and in need of verification through data. 51Harry E. Allison, Charles J. Zwick and.Ayres Brinser. "Recruiting and Maintaining a Consumer Panel" Journal of Marketing. 22 (April 1958) 377-390. 52Bucklin and Carman. p. 19. CHAPTER III THE MICHIGAN STATE STUDENT CONSUMER PANEL The Michigan State University Student Consumer Panel was established to study the income and expenditure patterns of Michigan State University students. The panel is a sponsored study, directed by Professor Gordon E. Miracle, Department of Advertising, and conducted over a two year period, 1967-69. General Panel Objectives and History The panel study concentrates on the general objective of describing student income and expenditure patterns. In particular, the panel focuses on several aspects of income and expenditures of Michigan State University (MSU) students; (1) the amount each student has available to spend, and the sources of this amount, (2) the amount spent in the greater East Lansing area, (3) the specific purchasing behavior with respect to day of week, product category, and specific store if in the East Lansing area, and (4) any predictable purchase patterns with respect to demographic, economic, or other variables. Administratively the panel study has been divided into a series of "Special projects" or sub-panels with respect to time of year, and to the various methodological manipulations that were made. The time of the year varied with the terms, the study running from Spring Term 1968 to the beginning of Spring Term 1969 in nine special projects. The methodolog- ical manipulations consist primarily of manipulations of the reporting period and of the communication method. 26 27 The reporting period manipulation involves the use of a "one day per week" design and an "every day per week" design. In the "one day per week" design the reporting period is shortened from the week as suggested by Lewis and others, to one day. Each respondent keeps the diary only for one specified day during the week, every week that the panel operates. The validation of this unique methodology will appear elsewhere, but preliminary analysis indicates the diary reports are very similar to those kept for the entire week (for that day). However, there is a distinct advantage in making the respondent's task easier, namely to keep him in the panel longer. Manipulation of communication method for purposes of this thesis focuses on two of the sub-panels. The first of these, denoted as Panel 1, consists Of a personal interview recruitment and a mail diary return, the second, or Panel 2, consists of a telephone recruitment and a telephone diary return. Panel 1 and Panel 2 Both of these panels ran for five weeks during the Summer Term 1968, and both used the one day per week design discussed above. Panel 1 received an introductory letter, and Panel 2 received an introductory letter plus a set of questionnaires, diary instructions and diary forms in a pre-recruitment mailing. The panels were drawn from the registrar's Summer Term enrollment lists using numbers drawn at random and comparing those numbers with the last digit in the six digit student number. The sample drawn from the registrar's data was then divided and assigned to either Panel 1, the personal interview recruitment, or Panel 2, the telephone recruitment. 28 Interviewers consisted of the 19 members Of the senior advertising research class taught during the Summer Term, who were given the inter- viewing assignment as a classroom assignment. The interviewers were not paid, but were provided with additional incentive in that their final assignments were a critique of the panel project. Interviewers were assigned approximately ten names for each panel: ten to be contacted in person, and ten to be contacted by telephone only. The interviewers were trained, including practice interviews, prior to conducting the interviewing. Data Collection Plan After prospective respondents received the introductory letter, and other materials in the case of Panel 2, the interviewers telephoned to make arrangements for the recruitment interview. They were instructed to make five calls before giving up the attempt to contact each respondent. In the case of Panel 1, the interviewers made an appointment for a personal interview. In the case of Panel 2, the interviewers either made the recruitment interview at this time or an appointment to call back and administer the recruitment interview at a later time. In the recruitment interview, in both Panel 1 and Panel 2, the Classification Questionnaire was administered. The Classification Questionnaire is primarily a demographic questionnaire intended to "classify" reapondents according to certain demographic characteristics. The interviewers also explained the project, explained how to keep the diary, what day to keep the diary, and when to begin keeping the diary. The Supplemental Data Questionnaire, a questionnaire designed to gather data relative to sources of income, was briefly explained and left with the respondent to be picked up at the post-recruitment interview. mesa .ma upsw=< mesa .NN sash mesa .ma sane mesa .HN-mH sane 29 moms .mH-m sane meme .8 sons mouse spasm Lo ecu coflumEuowcH human «0 asuxofim OOOLQOHOH cmwmm aow>houcm HmmuHcH 6cm unusumswomm economaoe mowumwa mCOmuosuumcH human muwmccoHummso mama HmucOeOHOQSm mufimccofiummac coHumOHuwmmmHu umuumq huouosuopucH "noxomm Ham: mua>auu< N Hpcmm swam GOWOOOHHOU sumo m manmh sesum mo ppm wcwacfiwom cwsuom human Hams manemo coHumEuomcH human mo annxuwm Hmcomumm Bow>hmucH HmHuHcH new mucmvcoomom uwsuumm wouumq mucuosvouucu «we: Nua>puu< a apnea 30 The post-recruitment interview was made the following week, in person for Panel 1, and over the telephone for Panel 2. During the post-recruitment interview the Supplemental Data Questionnaire was picked up along with the first Diary sheet. Any questions the reSpon- dent might have, or errors in filling out the diary that the interviewer would detect, could be answered or corrected. Panel 1 was left with envelOpes and instructed to return the remaining four diary pages through the campus mail. Panel 2 was called each week, with the panel member reading the diary page to the interviewer over the telephone. Panel 2 never re- ceived a personal visit from any panel representative, while Panel 1 received two visits. Follow-up telephone calls were made to approximately half of the panel members of both panels and asked questions relative to the ease or difficulty of keeping the diary and participating in the panel.) These calls were not made by the original interviewers. The follow-up question- naire, and all of the other questionnaires and materials are found in the appendix. The interviewers were also asked to keep a record of their hours and the number of calls and interviews they made. This record is the basis for the calculation of cost which will appear later. A careful record was also kept of panel members in terms of diary page receipt and questionnaire completion. The Diary The diary is the most important data collection form since it is the source of all Of the data relative to student expenditure and purchase patterns. Panel members were asked to record each purchase made on their 31 reporting day, recording the date, the product category, the store name, the store location, and the amount of each purchase. The panel was aimed at store patronage behavior rather than brand loyality or switching. Stores are recroded as Opposed to brands which are typically recorded in a consumer purchase panel. The product categories are explained fully in the Diary Instruction booklet and appear in abridged form at the beginning of each diary form, and at the top of every diary page. The product categories are briefly: (1) Books and Supplies, (2) Drugs, Medical Care, Personal Care, Cosmetics, and Toilet Articles, (3) Home Furnishings, (4) Laundry and Cleaning, (5) Magazines and NewSpapers, (6) Packaged and Fresh Foods, Groceries, Meats, Beverages, All Bakery Goods, Restaurant Meals, Liquor, Beer and Wine, (7) Recreation and Entertainment, (8) Rent, Utilities and Related, (9) Tobacco, Tobacco Products and Related, (10) Tuition, Board and Room, and Related Payments to MSU, (ll) Wearing Apparel, (12) Transportation Expense, (l3) Monetary Contributions to Church and/or Charity, (14) Miscellaneous Articles, and (15) Income and Other Money Received. A more complete explanation of these categories can be found in the Diary Instructions in the Appendix. These categories are both simple, and mutually exclusive. Little difficulty is encountered in determining which category a given purchase belongs in. The diary is both comprehensive, and simple to keep. CHAPTER IV SAMPLE MORTALITY The discussion of sample mortality will focus on three points: (1) the quantitative mortality at each of the three stages described 13y Bucklin and Carman, (2) a comparison of the characteristics of the :samples with the characteristics of the population, and (3) qualitative Inortality in terms of a comparison between the characteristics of panel tnembers and the characteristics of those who were lost. Quantitative Mortality: Contact and Response Rates Sample or panel mortality is most often discussed in terms of the