Research Panels in Consumer Behavior

Robert Ferber, University of Illinois
Linda B. Lannom, University of Illinois
[ to cite ]:
Robert Ferber and Linda B. Lannom (1981) ,"Research Panels in Consumer Behavior", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 238-244.

Advances in Consumer Research Volume 8, 1981      Pages 238-244

RESEARCH PANELS IN CONSUMER BEHAVIOR

Robert Ferber, University of Illinois

Linda B. Lannom, University of Illinois

INTRODUCTION

Most panel or longitudinal studies are undertaken to study some aspect of consumer behavior, such as purchase patterns over time or people's experiences moving into and out of the labor market. Such uses of panels are becoming more commonplace, and deservedly so, because they provide information about consumer behavior that only can be obtained by the panel technique.

At the same time, the growing popularity of the panel technique has led to a few attempts to set up panels to study the technique itself. This type of panel is much less glamorous, since no substantive question or policy issue is under investigation, and hence there is no expectation of dramatic results that, say, 46% of milk drinkers have seen the light and switched to scotch. The fact remains, however, that research panels serve a vital function in investigating the reliability of this technique and showing how the efficiency of these panel operations can be improved. Accordingly, the purpose of this paper is to discuss the objectives and the functions of research panels and to indicate the types of uses to which they can be put. In addition, we shall present some results obtained with one type of research panel that has been operated by the Survey Research Laboratory for more than ten years.

OBJECTIVES OF A RESEARCH PANEL

Before discussing the objectives of this type of panel, it might be a good idea to explain what we mean by a research panel. First, we should make clear that by a panel or longitudinal study we are referring to a series of reinterviews of the same sample on the same subject. We classify a study as a panel only if at least two reinterviews are involved, since the one reinterview case is more likely to be an experiment than a panel study. The reinterviews are usually made at regular intervals, and the same people and the same basic information are sought in each interview, though this does not preclude asking for additional information as well.

By a research panel we mean one where the primary focus is on some methodological aspect of a panel operation. This may include a panel that is to be used solely for experimentation with panel techniques or to investigate the cost or the feasibility of a large-scale panel by means of a pilot study. Research panels that we know of are invariably of the latter type. They may have objectives such as:

a.  To ascertain the types of problems likely to be encountered in that type of panel operation.

b.  To explore means of dealing with these problems.

c.  To obtain operating experience with such a panel over a period of time, especially, information on panel mortality, sampling biases over time, conditioning effects, and reactions to sensitive information.

Even though the focus may be on methodology, a research panel has to have a substantive focus as well. This is necessary to make the panel members happy, as well as the funding agency. Thus, it is hardly feasible to inform a sample of households that we plan to study how their answers or behavior may change over time as a result of membership in a panel. It is a lot easier, and more feasible, to tell them that we want to study how their saving and spending patterns change over time as their little monsters grow up and take over the household.

Moreover, this provides the opportunity to obtain useful information at the same time as valuable data are obtained on the more important methodological aspects. Also, it is easier for a funding agency to justify a methodological study when substantive results are to be obtained as well--policy makers do not easily comprehend spending tens of thousands of dollars to improve methodology with which they have little contact, whereas they are much more quick to see the value of a study that will give them more information on the feeding and control of the two-legged little monsters whom they run into all the time.

SOME EXPERIMENTAL CONSIDERATIONS

The planning of a research panel involves all of the usual considerations of a sample survey plus a few additional wrinkles. Thus, we still have to consider what population will be studied, how to select a proper sample in light of the given objectives, how to design the questionnaire, how to deal with possible response and nonresponse biases, and so on. In addition, however, some factors have to be kept in mind that do not usually arise in the ordinary pilot, or experimental, study.

For one thing, any experimentation should not be destructive in the sense of losing an appreciable part of the sample, at least not till the end of the study. For example, in a long-term panel operation, one cannot start out by asking, say, half of the sample for a complete listing of their assets and debts on the very first interview, to measure the effect of asking for highly sensitive information. While the results of such a procedure would certainly provide a good measure of the effects of asking for highly sensitive information at the beginning, this procedure will also very likely lead to the loss of an appreciable part of the sample, so that the sample size available for later observation and experimentation will be sharply reduced.

For these reasons, any experiments built into a research panel have to be planned keeping in mind their effect on the amount of information to be collected and the type of experiments to be carried out at a later time. One other consideration is that any methods used with the panel to improve response have to be planned in terms of their possible effect on the conditioning of the panel members and the types of information that will be sought in the future. Thus, any incentives or premiums have to be selected very carefully. For example, if the focus of a panel study is on the financial behavior of households, one can hardly give as an incentive a book on how to save and invest money. While it would be nice for some purposes to obtain an exact measurement of the bias due to this type of incentive, it is much the wiser approach from the point of view of study design to simply assume there will be some appreciable effect and to focus on other types of incentives for the purposes of the study. To be sure, there is no reason why some experimentation with incentives, or with other variables, cannot be built into a research panel, but it becomes virtually crucial that each of the experimental treatments be independent of the attitudes or behaviors to be studied as part of the substantive purpose of the study.

A third consideration is that if the research panel is to be operated over a number of years, every effort has to be made to maintain rapport with the panel members and to keep in contact with them. This moans that the interviews should not be so long as to irritate the panel members to the point of dropping out, that procedures have to be developed for keeping in contact with the panel members and finding out if they moved, and that where possible, other measures be used to maintain their interest, such as periodic reports or even a newsletter. Incentives can also be very effective.

Of course, ideally a research panel, like any survey study, should be based on a carefully thought-out set of objectives accompanied by a detailed set of data collection plans and analytical methods for meeting these objectives. In practice, this is hardly feasible with a research panel, and less so as the length of the panel study is increased. This is not to say that a long-term panel study may not have a set of specific objectives, and also general plans for meeting these objectives. However, conditions do change over time, resources may change, new ideas may arise from previous waves (if one is not careful...), and the priorities may be changed. Hence, while a general set of objectives is useful, it is also wise to stay flexible and be able to accommodate changes in objectives and procedures as they seem warranted.

Such changes are especially likely in a long-term research panel of the type that we shall describe very shortly. In such a case, there was the additional complication, that is not unusual for long-term panel studies, of a change in the funding sources, which can lead to changes in both the objectives of the study and the procedures.

In general, there seems to be somewhat of a tendency in the case of a research panel to place less emphasis on pure experimentation and to obtain more information on the techniques and methodology by means of observation and ex-post data analysis. This is not the place to indicate the various ways in which this may be done, except that we might note that this type of information can be obtained both internally--by studying the characteristics of the data--as well as by comparing characteristics and behavior of the panel with that of the population from which the panel was drawn.

THE ILLINOIS-BERKELEY PANEL

Background and Objectives

Three features make our consumer panels unique among studies of family economic behavior. The panel aspect is, of course, the primary of these. To date, 17 waves of data have been collected from the Decatur-Peoria panel, and 11 waves of data have been collected from the Chicago panel, creating two data sets which are rich sources of information about changes in family economic behavior and purchase decisions. Second, in these consumer panels, the family is not viewed as a single homogenous unit. Rather, it is regarded as being composed of individuals with separate backgrounds and personality characteristics which all influence family economic behavior to some extent. Finally, all couples in the panel are in the early stages of family formation and growth. This fact, together with the panel design, allows the collection of data from families while their economic and consumption habits are being formed and while many major life style events, such as the birth of children, migration, divorce, and remarriage are taking place.

Outline of the Panel Design

The consumer panels are two samples of young married couples who have been interviewed regularly since the time of their marriages. One of the panels consists of couples married in the summer of 1968 in Peoria and Decatur, Illinois. These sites were selected to avoid the complexity of a large city and to obtain respondents in sites which were reasonably self-contained economically and socially. At the time the panel was formed, the population of these two cities was between 100,000 and 250,000, their economies were primarily industrial, and they were relatively isolated from other major urban centers, as is still the case.

The Peoria-Decatur sample was selected from a list of 1,300 couples who married in Peoria or Decatur between June l, 1968, and September 30, 1968. Three sources were used to develop this list--county marriage license data from the three counties forming the Decatur-Peoria SMSA, newspaper accounts of new marriages, and records of clergy. From this list a sample of 400 newly married couples was selected, 150 in Decatur and 250 in Peoria. Addresses for these newly married couples were obtained by matching the last names with telephone directory listings or by contacting secondary sources, such as parents. Only couples who lived either in the Peoria or Decatur SMSA after their marriage were used. In addition, the husband had to be 30 years of age or less at the time of the marriage. The first wave of interviewing was completed In the fall of 1968, with 311 eligible couples completing the questionnaire.

When the Peoria-Decatur panel was four years old, the Chicago panel was formed, composed of young couples married in Chicago between June 1, 1972, and August 31, 1972. This sample was drawn from microfilm marriage records obtained from the Illinois Bureau of Vital Records. A sample of 1,O03 couples married in Cook County was selected from this list. Location of the newly married couples' addresses was accomplished as in Decatur-Peoria, using telephone directory listings and information supplied by secondary sources. Only couples who lived within the Chicago SMSA after their marriage were eligible.

Four other eligibility criteria were imposed in Chicago:

1.  The husband should be 30 years of age or less at the time of the marriage;

2.  The marriage should be the first one for both members of the couple;

3.  The couple could not own a home at the tine of the marriage;

4.  The couple's income should exceed $5,000 per year, unless one of them was a student.

Since it was difficult to find couples who met all of these criteria, the first criterion was relaxed to include couples with husbands who were 34 years of age or less at the time of their marriage. The Chicago panel, first interviewed in the late fall of 1972, contained 409 couples.

TOPICS INCLUDED

Economic Behavior

The major purpose of these panels is to collect data on the durable goods ownership and financial assets of these young couples. For this reason, on each wave couples are asked about their purchases of durable goods during the time between interviews and their intentions to purchase durable goods during some specified period in the future. At regular intervals, information on their stock of goods is updated. For many of these goods, data are available on such characteristics as the price paid, the brand name, on whose advice the purchase was made, and who made the decision to purchase. Extensive information is also collected on the couples' financial portfolios, saving and budgetary practices, and use of credit. Many of the questions are geared toward determining the role of each marital partner in the decision making process.

Data relevant to the economic personalities of the spouses are also collected periodically. Both husbands and wives respond to questions by which they can be characterized as economically minded, price conscious, or bargain seeking.

Subjective assessments of the country's current and future economic situation, including inflation, have also been gathered from the respondents. Intermittently, the subjects have assessed their own economic situation relative to that of their parents, their friends, their own past, and their anticipated futures. Data on household income have normally been collected once a year.

Labor Force Participation

Data on additional economic characteristics relevant to labor force participation, such as occupation, industry, part- or full-time employment, and involvement in a second job have been collected at each interview. The surveys have also included an extensive number of questions relevant to female labor force participation (most of these after Wave 9 in Peoria-Decatur and Wave 3 in Chicago). These included asking nonworking wives about their reasons for not working as well as their own assessments of their occupational prospects if they were to seek a job, and the cost of child care arrangements they would incur by working. Working women reported on the cost that they actually incurred while working as well as their reasons for working. All women responded to questions on their work plans after having children and in the more general future.

Background Characteristics

A series of questions on the respondents' background characteristics was included in the first wave for each of the panels. These included birthplace, occupation of father, number of siblings, parents' marital status, and whether parents were currently living. Current religion and religious background were sought on a later wave.

Aspirations and Goals

Periodic attempts have been made to measure respondents' aspirations and goals. These have included educational, occupational, and familial aspirations, as well as financial aspirations. One wave also included questions on respondents' aspirations for their children.

Children

At each wave, respondents indicated whether they had a birth during the interval between waves. On the tenth wave in Peoria-Decatur and the fourth wave in Chicago, the exact dates of all births were obtained. This provided a check on earlier answers and filled in missing data for a few cases. Respondents also reported at each wave whether they were expecting a child or in the process of adoption. After Wave 6 in Peoria-Decatur and beginning in the first wave in Chicago, respondents answered questions periodically on their desired and expected family sizes. Respondents have also been asked about birth control as well as about abortion. On various waves, an attempt was made to review the decision-making process surrounding fertility.

General Attitudes and Personality Characteristics

Throughout, attempts have been made to measure general attitudes toward life and the personality characteristics of the respondents. Questions regarding attitudes toward life in general and satisfaction with one's own life and marriage have been asked repeatedly. Among other characteristics, measures of experiment-proneness, timidity, conservativeness, and self-assuredness were constructed from responses to individual questions.

Sex Role Orientation

Both husbands and wives have responded periodically to questions concerning at what age and under what circumstances a mother/wife should work, the importance of job advancement, factors important in job selection, decision-making concerning employment of other family members, and work histories of respondents' mothers. Husbands and wives have also been asked about the division of labor within their homes as well as those of their families of origin.

Time Use

Use of time for household tasks and leisure activities was the focus of a number of questions on Waves 8, 9, 12 and 15 in the Peoria-Decatur study and Waves 2, 3, 6 and 9 in the Chicago study. A series of questions about vacation plans and habits was asked of the Peoria-Decatur respondents on the eighth wave.

Social and Political Attitudes

Some topics reflect the political and social climate of the times in which they were asked. For instance, respondents' attitudes toward electric cars, reusable containers, and organically grown food have been assessed. Questions on transportation were asked during the energy crisis of 1973-74 and again on the current 1980 wave of the panels. Attitudes toward protests by blacks, students, and truck drivers were measured at the time truckers were blocking the highways. At the same time, questions concerning violence on television and police violence were asked. During earlier waves, questions on media preferences (newspapers, magazine, television) and exposure were asked.

This list represents a fraction of measures taken during the study and should not be regarded as exhaustive. Many of the topics just mentioned were covered on the spouses' separate questionnaires, so that answers are available from both members of a couple. In addition, many of the topics were repeated, an answers are available to the same questions at different points in time and after different durations of marriage. The table provided as an appendix summarizes selected data currently available for the consumer panel.

RESPONSE EXPERIENCE

By June 1980, the Decatur-Peoria panel had been interviewed 17 times, and the Chicago panel 11 times. From 1968 until 1977, the aim was to interview both panels every six months. Beginning in 1978, the aim was to interview each of the panels once a year. Because of the vagaries of funding, it was not always possible to achieve these aims; thus, the actual intervals between waves of interviews varies between six and fifteen months.

The date of each wave of interviewing and the number of households completing interviews at each of the waves is shown in Table 1. From the first through the seventeenth wave, the total number of households in Decatur and Peoria dropped from 311 to and 206. The reduction in the number of cases was even more severe in the Chicago panel, which lost over half of its members from the first to the tenth wave, dropping from 406 to 194 cases. As expected, attrition was most serious during the earlier waves of the panel, due largely to couples refusing to participate or inability to locate couples who had moved.

We were concerned that such attrition represented an important source of bias in the consumer panel. To measure this bias, we conducted a follow-up study of nonrespondents in 1975, making intensive efforts to locate end obtain completed interviews from a sample of people who had left both panels. We found the dropouts of these consumer panels differed only slightly from the current panel members. The relatively high rate of attrition, particularly for the Chicago panel, appeared to have had very limited effects on the demographic structure of the panel. Significant differences between panel members and dropouts did appear in relation to expenditure characteristics, in particular to the ownership of various durable goods. However, even here, no discernible pattern of bias emerged.

EXPERIENCE WITH SENSITIVE INFORMATION

The focus of these panels is family financial decisions. Extensive information is collected on the couple's financial portfolios, saving and budgetary practices, use of credit, and ownership of life insurance. Many of these questions are highly sensitive, asking for the value in dollars of specific assets and debts. Our experience with item nonresponse for such items parallels our experience with attrition; that is, the rate of item nonresponse has dropped over the life of the panel. For example, item non-response to household income was .3 percent on the seventh wave of the Peoria-Decatur panel and .9 percent on the first wave of the Chicago panel. On the most recent wave of each of these panels, nonresponse to this item had dropped to .2 percent.

We believe that item nonresponse has dropped for several reasons. First, those most likely to refuse sensitive items have already left the panel. Second, over the years, the panel respondents have come to trust our guarantee of confidentiality. Periodically, we issue special reports to respondents, summarizing some of the findings of the study. In reading through these reports, respondents discover that we do, in fact, report their responses as aggregate estimates, and that it is not possible to identify individuals on the basis of what we report.

Third, in 1977, the method of data collection was changed from face-to-face interviewing to telephone interviewing. When we made this switch, we were concerned that it would depress response rates and possibly increase item non-response. In fact, it did not. We found that respondents were less reluctant to report dollar amounts when interviewed on the telephone, possibly because this form of data collection had the effect of making interviews more objective by insulating respondents from the interviewer.

EXPERIENCE WITH THE USE OF INCENTIVES

Because the panel members are asked to do a substantial amount of record checking during the course of an interview, we give them a small gift each time they ere interviewed. When respondents were interviewed face-to-face, the interviewers presented the gift at the conclusion of the interview. Now that respondents are interviewed by telephone, gifts are mailed in advance to each panel member. The monetary value of these gifts is small, ranging from $2 to $5 per couple. However, we feel the psychological value of the gift is important both for minimizing attrition and item nonresponse.

SOME USES OF THE DATA

Because the consumer panels are ongoing projects, analysis to date has been somewhat limited. Most of it has been done using data from the Peoria-Decatur sample, since the Chicago marriages are still of relatively short duration. The analysis which has been done has concentrated on two major areas--various facets of family financial behavior, and fertility and female labor force participation.

Very briefly, uses of the principal findings from these studies may be summarized as follows:

Family Financial Behavior

1. The studies support the idea that family financial behavior can be better understood if the family is treated as interrelated individuals rather then as a homogeneous unit. In their analysis of asset accumulation by the Peoria-Decatur couples, Ferber and Nicosia (1972) found that while education of both the husband and the wife influenced asset ownership, the effects of education for the wife tended to be in the opposite direction from that of the husband. In addition, the effects on asset accumulation of some personality variables and individual's priority on savings also tended to work in different directions for husbands and wives.

2. Ferber and Lee (1974a) found that who played the role of the so-called Family Financial Officer (FFO) did not appear to be determined by socioeconomic characteristics of the couple, but rather by savings attitudes and economic personalities. Most frequently, the couple acted jointly as the FFO; however, over time there was a pronounced shift toward the wife's assuming this role. The identity of the FFO did appear to influence economic behavior. For instance, when the husband was the FFO, the proportion of income saved was greater than when the wife alone or the couple acting jointly played this role.

3. Lee and Ferber (1977) found tendencies for assets to be greater when the husband and wife acted jointly as the FFO. They also found that when wives worked the family had less debt. However, the couples best off financially after five years of marriage in term of net assets were those who were venturesome enough to acquire substantial amounts of debt. In this sample, both debts and assets primarily related to the purchase of a home.

4. Ferber and Lee (1974b) have attempted to integrate lifestyle into their model, believing that lifestyle may provide an additional dimension beyond the socioeconomic role and personality variables studied to date. Operationalizing lifestyle according to time use, they classify couples into lifestyles on the dimensions of career orientation, pleasure orientation, and familial orientation. Findings indicate that lifestyle helps to explain why women work and whether couples purchase automobiles.

5. Several additional papers by Ferber and his associates (Ferber and Lee 1973; Liebman and Ferber 1977; Nicosia and Ferber 1976) have been concerned with the ability to predict purchases of certain durable goods from responses to questions on past purchases, satisfaction with current goods, and the likelihood of purchasing of goods in the future. The most recent paper of the three (Liebman and Ferber 1977) found that later actual auto purchases were related to earlier predicted and current auto purchases.

6. Anderson and Nevin (1975) found that demographic, socioeconomic, psychographic and other variables were relatively weak predictors of household life insurance purchasing behavior. They reasoned that this was due in part to the exclusion of life cycle change as an explanatory variable, since they used data for the first one or two years of marriage only. They did find that six variables were significant in explaining the amount of life insurance purchases.

7. Lee and Ferber (1980) in a further study of life insurance acquisition found that three variables were significant in explaining the type of life insurance purchased. The purchase of term insurance was found to be much more likely in households where the net worth was large; the wife purchased term insurance before marriage; and the insurance agent did not influence the couple's decision.

Fertility and Female Labor Force Participation

8. Several papers using data from the consumer panels have concentrated on the determinants of fertility behavior. In "Relative Status and Fertility" Crimmins-Gardner and Ewer (1978) tested the relative income hypothesis that fertility is affected by the relationship between desired and actual living levels. Relative status was measured as the difference between the husband's occupation and the occupations of both the husbands' and the wives' fathers.

Also, each spouse compared his (her) family financial situation at the age of 16 to the couple's financial situation at marriage, in the seventh year of marriage and with the expected situation in the seventeenth year of marriage. When these measures of relative status were related to fertility, only the wife's subjective assessment of relative status showed a consistent and significant pattern. Wives who felt they were or would be better off than their parents, want more children than wives who compare their own situations unfavorably to their teenage years. Hypothesis that people with certain backgrounds which emphasized children over material possessions have children earlier relative to the time they acquire material goods (when income is controlled) than people without these backgrounds. Occupation and religion were the two background characteristics examined. However, the results showed that occupation and religion have little effect on the timing of the first child, the purchase of durable goods, or the relative timing of these two acquisitions. Income does affect when people acquire goods, both absolutely and relative to children. Couples with higher incomes acquire goods absolutely earlier, and earlier relative to the birth of the first child. This last relationship is the result of both earlier goods acquisition and later births of first children in couples where the wife works.

10. A panel design is particularly appropriate for investigating the causal dynamics underlying the negative relationship between working and fertility. Data from the first seven years of the Peoria-Decatur panel were used in a path analysis to disentangle the effects of husband's income, wife's education, wife's age at marriage, family size, and wife's employment and subsequent time intervals (Ewer, et al., 1979). The results indicated that during the early stages of marriage and family formation the presence of young children exerted a strong negative effect on wife's employment. In contrast to this effect of fertility on employment, the effects from employment to fertility are less consistent.

11. In an additional study, only wife's labor force participation was used as the dependent variable (Crimmins-Gardner and Ewer 1978), to explain factors affecting employment among young mothers. In addition to the usual determinants of working, expenditures connected with working and the emotional cost of child care were important determinants of the labor force behavior of young mothers.

12. Bagozzi and VanLoo (1978a, 1978b, 1979), have begun to develop a theory of fertility as consumption, examining tree classes of socioeconomic determinants--economic variables and social structure variables. The socioeconomic determinants are hypothesized to act as exogenous variables, either facilitating or constraining psychological and social decision-making processes within the family. These latter processes are posited to be the primary endogenous variables affecting fertility.

Tests of the overall theory were performed on three separate samples using a structural equation methodology. In general, fertility decisions were shown to be affected directly by the social psychological processes within the family and only indirectly by socioeconomic constraints, as predicted.

IF WE HAD TO DO IT OVER

Recounting these experiences and results leads us to raise a final question that is likely to be of general interest, namely, if we had the opportunity to do this panel study over again, would we, and how would we change things?

First, would we do it over again? Certainly--the topic is fascinating, the study design is full of challenges, and the results seem to be both very useful and personally rewarding. On the other hand, we would also certainly want to make various changes in the study design. Some of these changes are pretty obvious, such as having a larger sample size and interviewing panel members every six months on a regular basis, something that we could not always afford. Indeed, interviewing the panel members in a study of this type every six months initially, say for the first five years, and every year thereafter would seem just about optimal, based on our experience.

Some of the other changes are perhaps less obvious, such as the following:

1. Even with separate interviewing of both spouses (a highly desirable approach for data collection purposes), we would have started using the telephone approach much earlier than we did. Our experience, as well as that, for example, of the Health Interview Survey of the U.S. National Center for Health Statistics, indicates that such simultaneous interviews can be obtained very efficiently by telephone. Personal interviews would still be used the first couple of times, and would also be used where families ere not reachable by telephone, but the principal reliance would be on the telephone approach.

2. We would seek more powerful methods of maintaining panel member cooperation. This could range anywhere from sending more letters and reports between interviews to introducing an incentive system similar to that of the commercial panels where the rewards for remaining in the panel increase geometrically with the length of tine. However, there would certainly be problems in doing so, given the University sponsorship and likely limitations on funds.

3. Insofar as possible, we would develop at the outset a clear set of objectives for the types of decision problem that we wished to study, or hypotheses to test, and develop a long-run data collection plan based on such objectives. Unfortunately, this is more easily said than done. Because funding sources change, there is always uncertainty regarding future funding, and even with a fixed set of objectives, one's ideas how to best obtain the necessary data change with time and with experience.

4. Last but not least, we would take our programmers and systems analysts, lock them in a room, and feed them only mashed potato sandwiches and warn water (or, worse yet, milk), until they came up with a simple and economical means of putting these data in a computer file that is amenable to analyses both over time and over space. Getting a manageable file that would permit the many uses to which panel data could be put is one of the most frustrating aspects of a panel operation. In fact, we are not sure that we would put the mashed potatoes into the sandwiches' until the programmers were close to a reasonable solution of this problem.

Thank you.

TABLE 1

INTERVIEW DATE, COLLECTION METHODOLOGY, AND NUMBER OF HOUSEHOLDS COMPLETING INTERVIEWS AT EACH WAVE:  CONSUMER PANELS OF YOUNG MARRIED COUPLES

APPENDIX

SELECTED DATA CURRENTLY AVAILABLE-CONSUMER PANELS OF YOUNG MARRIED COUPLES

----------------------------------------