Using Unobtrusive Methods in Consumer Research: an Assessment
ABSTRACT - The study discusses the advantages and disadvantages of the survey approach. It is hypothesized that observational methods sometimes constitute reasonable and powerful substitutes to traditional survey methods. Depending on the environment of the phenomena under study, unobtrusive approaches may even outperform traditional survey methods. Non-reactive approaches seem especially appealing in situations, where instrumental effects threaten to bias findings. The empirical study investigates whether it is possible to estimate readership of trade circulars (sales flyers) by uncovering fingerprints on the pages. Samples of issues collected at a local recycling center were appropriately analyzed by an expert. According to the pilot study involved, it is possible to measure readership by analyzing fingerprints. However, an estimate thus generated can probably only serve as a conservative prognosis (lower bound) with regard to the true but unknown readership of a flyer. The paper finishes with a discussion of future studies using unobtrusive methods in consumer research.
Marcus Schmidt and Niels Krause (2001) ,"Using Unobtrusive Methods in Consumer Research: an Assessment", in E - European Advances in Consumer Research Volume 5, eds. Andrea Groeppel-Klien and Frank-Rudolf Esch, Provo, UT : Association for Consumer Research, Pages: 180-186.
The study discusses the advantages and disadvantages of the survey approach. It is hypothesized that observational methods sometimes constitute reasonable and powerful substitutes to traditional survey methods. Depending on the environment of the phenomena under study, unobtrusive approaches may even outperform traditional survey methods. Non-reactive approaches seem especially appealing in situations, where instrumental effects threaten to bias findings. The empirical study investigates whether it is possible to estimate readership of trade circulars (sales flyers) by uncovering fingerprints on the pages. Samples of issues collected at a local recycling center were appropriately analyzed by an expert. According to the pilot study involved, it is possible to measure readership by analyzing fingerprints. However, an estimate thus generated can probably only serve as a conservative prognosis (lower bound) with regard to the true but unknown readership of a flyer. The paper finishes with a discussion of future studies using unobtrusive methods in consumer research. INTRODUCTION: PROS AND CONS OF OBTRUSIVE METHODS Historically, the survey approach has been the prevailing technique for gathering marketing research information, while observational methods have been applied far less frequently. Due to several reasons, both the academic community and the commercial industry apparently prefer the questionnaire to observational investigation. Why is this so? First, when using a survey the researcher has full control of the experimental design. Second, the questionnaire can be structured properly. Third, the method is well-established. Fourth, responses are easy to analyze and thus results are soon at hand. Fifth, a wide array of powerful methods for testing reliability and validity is available to the researcher. According to Kellehear (1993) "There is today ... a simple and persistent belief that knowledge about people is available simply by asking." There is little doubt, though, that traditional survey methods also have drawbacks. The cardinal conundrum is this: Can we trust what the respondent tells us she/he has done, does or intends to do? Is her/his reporting of past, present and future behavior at all reliable? In most situations it is believed that responses are trustworthy and correct. However, depending on the research environment and the phenomena under investigation, this may not be so. Over- and underreporting as well as back-/forwards telescoping can bias results. Sometimes, i.e. when sensitive topics are under investigation, findings will be useless due to lack of external validity (Sudman and Bradburn 1983). In research settings dealing with low-involvement buying behavior, respondents - due to questionnaire ambiguity - may simply not be able to provide reliable answers to the researchers questions (Green, Tull, and Albaum 1988, 261-75; Ackoff, Gupta, and Minas 1962, 179). For discussions about nonsampling errors see Bradburn and Sudman (1988, 185-91) and Noelle-Neumann and Petersen (1996, 86-92). In a study by Shimp and Kavas (1984) the authors raise serious doubt about the criterion or construct validity of the self-reported approach: "
The study discusses the advantages and disadvantages of the survey approach. It is hypothesized that observational methods sometimes constitute reasonable and powerful substitutes to traditional survey methods. Depending on the environment of the phenomena under study, unobtrusive approaches may even outperform traditional survey methods. Non-reactive approaches seem especially appealing in situations, where instrumental effects threaten to bias findings. The empirical study investigates whether it is possible to estimate readership of trade circulars (sales flyers) by uncovering fingerprints on the pages. Samples of issues collected at a local recycling center were appropriately analyzed by an expert. According to the pilot study involved, it is possible to measure readership by analyzing fingerprints. However, an estimate thus generated can probably only serve as a conservative prognosis (lower bound) with regard to the true but unknown readership of a flyer. The paper finishes with a discussion of future studies using unobtrusive methods in consumer research.
INTRODUCTION: PROS AND CONS OF OBTRUSIVE METHODS
Historically, the survey approach has been the prevailing technique for gathering marketing research information, while observational methods have been applied far less frequently. Due to several reasons, both the academic community and the commercial industry apparently prefer the questionnaire to observational investigation. Why is this so? First, when using a survey the researcher has full control of the experimental design. Second, the questionnaire can be structured properly. Third, the method is well-established. Fourth, responses are easy to analyze and thus results are soon at hand. Fifth, a wide array of powerful methods for testing reliability and validity is available to the researcher.
According to Kellehear (1993) "There is today ... a simple and persistent belief that knowledge about people is available simply by asking." There is little doubt, though, that traditional survey methods also have drawbacks. The cardinal conundrum is this: Can we trust what the respondent tells us she/he has done, does or intends to do? Is her/his reporting of past, present and future behavior at all reliable?
In most situations it is believed that responses are trustworthy and correct. However, depending on the research environment and the phenomena under investigation, this may not be so. Over- and underreporting as well as back-/forwards telescoping can bias results.
Sometimes, i.e. when sensitive topics are under investigation, findings will be useless due to lack of external validity (Sudman and Bradburn 1983). In research settings dealing with low-involvement buying behavior, respondents - due to questionnaire ambiguity - may simply not be able to provide reliable answers to the researchers questions (Green, Tull, and Albaum 1988, 261-75; Ackoff, Gupta, and Minas 1962, 179).
For discussions about nonsampling errors see Bradburn and Sudman (1988, 185-91) and Noelle-Neumann and Petersen (1996, 86-92). In a study by Shimp and Kavas (1984) the authors raise serious doubt about the criterion or construct validity of the self-reported approach:
"The use of retrospective self-reports rather than actual ... usage behavior is a notable research limitation. The extent to which self-report data mirror actual ... usage is problematic. A field study was performed to validate the accuracy of self-reported data. Research assistants were positioned in two grocery stores to observe unobtrusively actual coupon usage behavior. A sample of 205 shoppers, who were ignorant that their behavior had been observed, later received a mail questionnaire which included questions regarding self-reported coupon usage. Self-report data from the 146 responding households were correlated with their previously observed coupon redemption behavior. A statistically significant though modest correlation was obtained (r = 0.32...). The absence of a stronger relationship is due, on the one hand, to the inherent fallibility of self-reported data, on the other hand, to imperfections in the validation procedure." (800)
NON-REACTIVE TECHNIQUES (I.E. OBSERVATION)
A quarter of a century ago Ray (1973, foreword) argued that there has been "an overdependence on interview (i.e. obtrusive) measures in marketing research." In a recent review article on the future of marketing research Malhotra, Peterson, and Kleiser (1999) provide recommendations concerning future methods for collecting marketing research information: "The challenge for ... researchers will be to use ... technologies ... in natural settings and in less-intrusive ways."
Thanks to recent technological improvements observational or tracing methods are increasing in popularity. This development has been facilitated by advances and breakthroughs in a variety of fields like data warehousing, data mining, and neural networks. Since the mid-eighties retail scanning has been has gained widespread use in marketing and consumer research. See Walters (1988, 1991), Walters and MacKenzie (1988), Kumar and Leone (1988), Karande and Kumar (1995), Mayhew and Wiener (1992).
"Classical" examples of empirical unobtrusive studies are (Webb et al. 1966):
$The wear of floor tiles in a museum indexed by the replacement rate was used to determine the relative popularity of exhibits
$The setting of car radio dials brought in for service was used to estimate share of listening audience of various radio stations
$Cigarette buts collected after a football game were treated as indicator of market share of selected brands
In other cases household garbage and toilet graffiti has been subject to detailed content analysis.
Advantages of unobtrusive methods, according to Kellehear (1993, 5-8) are:
$They study actual rather than reported behavior
$Are regarded safe (discrete procedure)
$Easy repeatable (re-checking is possible)
$Non-disruptive, non-reactive (non-involving)
$Research access is very easy (no cooperation of others is necessary)
$They are often inexpensive
$Constitute good source for longitudinal analysis
While disadvantages are (same source):
$The original records may be of poor quality or distorted
$Records are seen from the point of view of the stranger (decontextualizing)
$Intervening variables may distort the picture
$The recording as such is selective and may be biased
$They may result in an over-reliance on a single method
$The application (interrogation) range usually is rather limited (narrow focus)
DISTRIBUTION AND DISCARDING OF CIRCULARS
DISTRIBUTION AND DISCARDING OF CIRCULARS
In Figure 1 we have tried to model how circulars / flyers are distributed to and afterwards discarded by Danish households. The Danish Society consists of about 2.2 million households. Basically, an advertiser (producer or retail chain) wanting to forward his circular to a broad audience, can choose between two big distributors and several small ones. The two big distributors - the Danish Postal Service and a commercial player (Forbrugerkontakt) - are able to deliver advertising messages to both the majority of households and to geographically specified target markets (other segments are difficult to target because of legal restrictions ensuring privacy of individuals). Due to cultural habits and legal considerations the overwhelming majority of circulars distributed to Danish households are unaddressed, while in other countries circulars are either addressed to "Resident of X-Street Number Y" or even to "John Doe, X-Street No. Y".
While the Danish Postal Service is obliged by law to deliver material to every Danish household, its main competitor (Forbrugerkontakt) does not distribute circulars in rural districts with a low population density. Therefore, this private company only covers 80% of households (1.7 mio).
According to a recent poll, more than two thirds of all circulars that households receive are recycled. While 20% of them end op in a container reserved for circulars and magazines, about half of them are put into a container allowing for all forms of paper, especially newspapers. Approximately one out of every six flyers are put into a disposable bag and thus mixed with ordinary garbage. Finally, five percent burn up in a tiled stove (See Figure 1).
EVALUATING READERSHIP OF ADS BY ANALYZING FINGERPRINTS: TECHNICAL CONSIDERATIONS
In this section we describe how the empirical sample was collected and subsequently analyzed.
First, a sample of 20 circulars was extracted randomly by using gloves (fear of contamination, self-integrity considerations) at a local recycling center (table 2). Fortunately, the center has separate containers for newspapers and for trade circulars (and weeklies). The total sample involved 30 circulars. They were gathered in January 1999 by use of different procedures. Ten "fresh" items (table 1) were collected immediately after having been studied by a few nonrandom respondents, selected by one of the authors. One of the authors was personally observing and supervising the event, thus ensuring that every page of the involved circulars was touched via fingertips by the individual respondent and by no one else.
Next, the sample was sent to and afterwards analyzed by one of the authors who works at the headquarter of The Danish Commissioner of Police. The Commissioners office includes a department for analyzing fingerprints. This department contains a small office with employees that specialize at locating fingerprints on paper. Usually these experts look for fingerprints on bad checks, forged banknotes etc. However, they do have the expertise to analyze a wide array of paper, including trade circulars for human fingerprints. Fingerprints appear because people sweat. 98% of human sweat consists of water and 2% of amino acid. The water must evaporate prior to analysis, since only the "pure" amino acid is traceable. Sweating is an individual characteristic. Typically, fat, young, temperamental, and nervous individuals sweat more than others. Likewise sweating correlates with season: People are sweating much more on a hot summer day as compared to a cold winter evening. To complicate matters, individuals with (1.) comparable physical and psychological characteristics, and (2.) being exposed to same conditions (temperatures) tend to generate different amounts of sweat.
Analyzing fingerprints on paper is a rather complicated procedure: First, all pages of a circular need to be separated. Second, every page has to be dipped into a liquid called ninhydrin. Third, the wet page must be dried in a high tech device resembling a big microwave oven. Finally, each page has to be scrutinized for fingerprints appearing on the page (A magnifying glass or even a microscope may facilitate and support the process of inspection). Unlike in criminal investigations, it is not necessary to perform a match concerning the uniqueness of an individual fingerprint. Based on prior behavioral studies and experience, the expert has a reasonable idea concerning where precisely on a page to look for fingerprints (far left, middle and bottom of left page and far right, middle and bottom of right page). [Since the procedure described here was not in use (even within criminal investigations) before the late nineteen-seventies, early efforts to use the technique within marketing research were deemed to fail. However, technological breakthroughs in measurement of fingerprints has changed this.]
What causes interest is whether the page contains a fingerprint or not. It is assumed that a person has been exposed to a page (and to all other pages!), provided that at least one valid fingerprint is found on a page. This corresponds to the Opportunity To See (OTS) concept - the prevailing definition of readership being used in interview-based surveys of magazine and newspaper readership: If the respondent reports readership of a specified media, it is automatically assumed that she/he has been exposed to all pages (and imbedded advertisements). [While the rationale of this assumption may be questioned, indeed, it will not be addressed in this paper.]
Given the setup chosen it was not possible to make assumptions concerning the characteristics of an individual fingerprint: Did it belong to a small girl or to a grown up male? Additionally, fingerprints appearing on the pages of a given circular may belong to different persons (of a household). According to the expert these problems might have been included in the research design. However, including this feature would have made the analysis much more complicated. Moreover, findings derived from such detailed analysis will be of a speculative nature.
In most cases fingerprints only appear a few times across a trade circular. If the circular consists of, say, 32 pages, then one typically will uncover only 2-4 fingerprints across all pages.
Furthermore, the frequency of identified fingerprints varies inversely with the number of pages of the individual trade circular: More fingerprints will be found on the first pages as compared to the last ones. Nevertheless it is reasonable to assume that the person who has left a few fingerprints, say, on the first four pages indeed has been flipping through the whole circular - although no traces are to be found on the remaining 28 pages.
For methodological reasons the front and rear page of each circular were excluded from analysis since they may have been touched (contaminated) by the messenger (an errand boy on a bike) or by the person of the household who carried the circular to the recycling center.
Unlike his wife (her husband), there is a chance that he (she) himself (herself) never read it. However a figure that includes front and rear page is recorded in tables 1 and 2 (numbers in brackets) for illustrative purposes (one should keep in mind that they are not valid for scientific purposes).
1. Non-random test sample
The study reported on here is a pilot study and is part of a far more comprehensive project being under preparation. The purpose of the pilot study has been to generate knowledge and experience concerning the measurement of fingerprints on circulars.
Table 1 displays the findings of the analysis of fingerprints with regard to five non-randomly selected respondents. ID-numbers 01/02, 03/04, 05/06, 07/08 and 09/10 refer to the same respondent, respectively. Each respondent was exposed to two circulars. Person 01/02 refers to one of the authors. 01 was read immediately after spending 20 minutes in a sauna. 02 was studied after having walked around for 20 minutes outside a Helsinki hotel in early January only wearing a T-shirt (-15oC).
The supposition that a person generates more sweat and thus more finger-imprints when he sweats as compared to when he freezes, was confirmed. However, the "sample" consisted of one (!) person. Numbers in brackets in the far right columns of table 1 relate to the total number of fingerprints traced on the circular, including front and rear page. Unfortunately, these fingerprints could relate to someone different from the reader(s) - see above. Therefore, these gross-estimates were not used for statistical and computational purposes. Two respondents (03/04 and 05/06) were exposed when the room temperature was 21oC, while the other two (07/08 and 09/10) were exposed at 25oC.
Findings based on the Empirical Sample:
$One respondent (03/04) provided no valid fingerprint.
$Four circulars (03, 04, 08, 09) contained no valid fingerprints
$Two circulars (04, 08) had no fingerprints whatsoever (even not on front/rear page)
Concerning four respondents 03/04, 05/06, 07/08, 09/10 we can make the following estimates: In total respondents were flipping through 112 pages. For each of the eight circulars we must detract the front and the rear pages. Thus, we end up with 96 valid pages. Across these pages the expert found seven fingerprints. Stated differently, 7% of all pages analyzed contained fingerprints. If we allow for the front and rear page to be included the figure rises to 11%. [In the present pilot study (involving IDs 03-10) doing so might seem justified, since the sample was drawn from a stable of new circulars, supplied by the retailers. Probably, they were solely touched by the researcher, who only touched them while wearing gloves.]
This computation can be repeated once for every respondent. The number of fingerprints varies considerably from respondent to respondent. For instance Respondent 05/06 is much more successful with regard to generating traces as compared to 03/04.
2. Random Sample
Table 2 displays results concerning the twenty randomly selected circulars. We possess no background information on the readers of these circulars. The 20 circulars were randomly selected across two days at the local recycle station and drawn from a container reserved for circulars and magazines. On the first day of data gathering the sky was clear and the weather was rather cold (+4oC, +39oF), while it was raining on the second day (+7oC, +45oF).
Although the temperature on the day of collecting the items has little influence on the quality of the fingerprints, wet conditions may deter traces. What really counts is the temperature and the conditions prevailing in the environment of the person while she/he is flipping through the pages of the circular. Doing it in front of a fireplace will obviously generate more sweat as compared to doing it in a cold kitchen.
FINGERPRINT-ANALYSIS OF FIVE NON-RANDOMLY SELECTED RESPONDENTS
The sample analyzed here was collected during wintertime when people do not sweat much, ceteris paribus. So, the technical conditions for identifying fingerprints were far from ideal.
Table 2 also displays the highest as well as the average daily temperature at the day the circulars were received by the households. [The day of reception could be positively identified since one of the researchers possesses an archive where all locally distributed circulars across fifteen years are registered and the day of reception is notified.] Next, the temperatures of the corresponding days were collected from the nearest meteorological office. The office performs measurements approximately 20 kilometers from the city were the sample was drawn.
The sample consisted of 860 pages and 55 valid fingerprints were identified. 60% of the circulars identified (12 out of 20) contained at least one valid fingerprint. We assume that the 60% is a lower level or bound with respect to readership. Regarding the remaining eight circulars (seven of them contained no fingerprint at all) we cannot state that they have not been read. We can only say that no fingerprints were traceable. It is remembered that respondent 03/04 in table 1 did not provide any valid fingerprints, although one of the authors observed her while performing the task. Note also that no valid fingerprints were identified on ID 30. However, it must have been studied since someone had encircled the price of a product with a pen.
Once more we observe significant variation across circulars. IDs 17, 21, 23, 25 and 29 produced lots of fingerprints, while others like 11, 16, 20, and 24 only left a single fingerprint. We can just speculate about the underlying reasons for this (see above).
Looking at table 1 it seems to make sense exploring whether there is correlation between gender, age, weight, height etc. of respondents and the propensity to produce fingerprints on ads. This phenomena can be investigated using a multiple regression analysis. But the test sample is very small (four respondents), thus making it impossible to identify significant correlations between, say, weight + age + room temperature and the propensity to produce fingerprints.
Concerning table 2 we do not have any background information on the respondents. However, we do have more observations and are thus able to formulate a tentative regression model, based on the data provided in table 2:
E(valid fingerprints)circular j = Comax, day j + Cox, day j +
Eof pages circular j + paper format j
Unfortunately, the multiple coefficient of determination is quite low (.23) and not significant due to the still small sample (20 observations). None of the four explaining variables have parameter estimates significantly different from 0. There is, however, a slightly positive correlation (.11) between the dependent variable (number of fingerprints) and the number of pages of the circular (which is no big surprise and of little interest).
Alternatively to using regression one could define a two group discriminant model where the classification variable could be paper quality (magazine versus newspaper quality) and the explaining variables being the remaining columns of table 2. Results are shown in table 3.
The significant value with regard to the number of pages makes sense and can be explained by typographical reasons (desk top publishers prefer the magazine format as the number of pages grows). We cannot explain, however, why there is a significant F-value concerning the corresponding temperatures. We think these significances are spurious and due to the small sample size. The expected correlation between paper quality and fingerprints was not significant.
Figure 2 displays the results of a multiple correspondence analysis based on the data in table 2. However, some manipulations have been performed prior to analysis. First, circulars that contained no valid fingerprint were coded as "Fingerprints0" while circulars containing at least one fingerprint were coded as "Fingerprints1+". Second, circulars having fewer pages than the average number of pages were coded "PagesFew", whereas circulars with above an average number of pages were coded "PagesMany". Finally, the two measures on temperature were coded accordingly as Co-maxL, Co-maxH,Co-aveL and Co-aveH respectively, with the average value on each of the two measures being the cutoff value between the H(igh) and L(ow) level. Technically speaking, the two temperature variables of table 2 were transformed from interval scales to categorical scales while the variables measuring the number of pages and fingerprints were altered from discrete ratio scales to categorical scales (the variable measuring paper quality was left unchanged since it was dichotomous in its raw format). To summarize, four of the five variables were "categorized" for analytical purposes.
FINGERPRINT-ANALYSIS OF TWENTY RANDOMLY SELECTED TRADE CIRCULARS
When inspecting Figure 2 we get some clues concerning association: First, appearance of valid fingerprints ("Fingerprints1+") is associated with high temperatures (Co-maxH Co-aveH) while the absence of fingerprints ("Fingerprints 0") is somewhat related to low temperatures (Co-maxL Co-aveL). Second, newspaper quality is associated with few pages while magazine quality is related to many pages. These two findings are of little surprise. However, the optical proximity of magazine-quality paper and the appearance of valid fingerprints ("Fingerprints1+") was not expected and is somewhat contra-inductive. It is not quite in harmony with what is hypothesized elsewhere in the paper. So far we have no explanation to this finding but we assume that it can be attributed to random error due to the modest sample size. [Note also that the sample contained 14-circulars of magazine quality and only 6 of newspaper quality. Moreover, our somewhat obscure and arbitrary "categorical recoding" does not properly take into account that the newspaper quality circulars on average contained few pages and relatively many fingerprints (and vice versa concerning circulars of magazine format). Finally, our MCA object space has a "forced" low dimensionality: The two first principal inertias only account for about 70% of the across-all-five possible dimensions Chi-Square value.] [The interpretation is based on points found in approximately (roughly speaking) the same region of the space. Distances between points do not have a straightforward interpretation in MCS. The geometry of MCA is not a simple generalization of the geometry of simple correspondence analysis (Greenacre and Hastie 1987, Greenacre 1988). Even in simple correspondence analysis comparisons between row point and column points is not valid (Greenacre 1989).] One of the authors (the fingerprint-expert) - due to his prior experience - assumed that it would be somewhat easier to find fingerprints on paper of newspaper quality. Table 2 provides some support for this hypothesis: While the 108 pages of newspaper quality produced 13 valid fingerprints (.12 per page) the 752 pages of magazine quality generated 42 fingerprints (.056 per page). Since the ratio between the two figures is 2.14 (.12/.056) the newspaper quality is approximately twice as successful with regard to identifying fingerprints. This result, too, is preliminary and needs to be confirmed by a bigger sample (research on this issue is preparation).
DISCUSSION AND FUTURE RESEARCH
Our analysis of fingerprints indicates that at least 60% of the circulars must have been studied by at least one individual. The remaining 40% may (1.) have been read by one or more individuals who have left no trace of the task, (2.) they may have been discarded by the household unread. (3.) There is a chance, however, that the circular never made its way to the prospects (See Figure 1): It is usual that the person distributing the circulars is equipped with too many circulars as compared to households or addresses en route. Since these printed ads typically have a life span of a single week, they are to be compared with perishable goods.
Consequently, the messenger - either a postman or a teenager on a bike and/or with a box barrow - will carry the remaining undistributed circulars to the recycling center and throw them into the container. Usually such circulars are easy to identify because they show up in the container as stacks of the same circulars, whereas other circulars are distributed randomly across the container (in the present study care was taken to circumvent this problem). According to annual polls on readership of circulars, conducted by the Danish Gallup company, approximately 80% of persons interviewed report readership on circulars for supermarkets and related retail stores. The readership of specialty retailers like fashion-, toy-, and hardware stores is slightly lower. Only about or less than every second person reports to read circulars for retailers selling prducts like TVs, computers, and photographic equipment. Our study does not contradict with these findings (11 of the 20 issues originated from supermarkets and only 2 did not contain fingerprints).
DISCRIMINANT ANALYSIS OF THE DATA OF TABLE 2 WITH PAPER QUALITY BEING THE CLASSIFICATION VARIABLE
MULTIPLE CORRESPONDENCE ANALYSIS OF FINGERPRINTS AND RELATED "CATEGORIZED" VARIABLES
Presently research is carried out based on a sample of several circulars that were distributed during very hot days in May and early June of 1999. During some of the days the temperature was above 30 Co. At the same time care was taken that the days, on which the circulars were received by households, were outside (just prior to) the normal Danish industrial holiday period. We assume that by choosing this research setting we improve our possibility with regard to identifying fingerprints. Most people must have been sweating on the day they received the circulars, while only few of them might have left their home for holiday (when the probability of not reading the circulars is greater). The fingerprint-analysis of these data have just been finished and presently the data are analyzed for construct and criterion validity using an array of multivariate statistical methods.
Reports on this bigger study will be published in spring 2002.The work-in-progress study is based on 120 circulars. While this may still seem to be a small sample, analyzing the material for fingerprints involved separating, treating and individually scrutinizing of approximately 5000 (!) pages. Consequently, the technical analysis has necessitated more than a year. Preliminary results of this large-scale random sample seem encouraging. One of the strengths of our approach is that we can estimate readership across "brands" of circulars. For instance, it seems that circulars distributed by some retailers are more popular than those distributed by others. So far, measurement of circulars across retailers has not been possible to cover by using the self-reported approach. Due to limitations linked to the self-reported approach, data can only be gathered on a "generic" level. (A Questionnaire by Gallup in Denmark only differentiates between circulars originating from "Supermarkets", "Males Clothing", "Toy Stores", "TV/Radio Stores", "Stores on Furniture Equipment" etc.)
In a follow-up study it is planned to combine the purely observational method reported on here with interviews. In a modified research setting we expect to interview respondents when they approach the paper container at the recycle center. First, they will be asked whether we are allowed to "inherit" their collection of circular for further study (of fingerprints etc). Next, they will be asked a couple of screening questions on their background. Finally we will ask the person for his phone number and ask him/her if we are allowed to call back at a later point of time.
In a different and unrelated study on of the authors is coding and analyzing a sample of five hundred shopping lists, discarded by consumers and found at parking lots, in garbage boxes etc. outside grocery stores (i.e. at supermarkets and discount stores). We expect this study to shed some light on topics like:
$How do consumers plan their shopping trip?
$How many products appear on the average list (and what is the variance)?
$Which product categories are most frequently mentioned?
$To which degree do consumers note brand names ("Jacobs") and not just product categories ("Coffee")?
$How man consumers do explicitly note prices and quantities?
$Is it provable or probable that the shopping list is directly influenced by the appropriate circular distributed by the store, informing about products on sale weekt.?
The research approach used in the paper presented here - to the best of our knowledge - has not been seriously investigated by marketing scholars up to this point of time. However, our approach in some regards resembles the analysis of consumer garbage (Ritenbaugh and Harrison, 1984, Cote, McCulloch, and Reilley, 1985) and may be seen as a continuance of this research tradition.
Ackoff, Russell L., Shiv Gupta, and J. S. Minas. 1962. Scientific Method: Optimizing Applied Research Decisions, New York: John Wiley.
Bradburn, Norman M. and Seymour Sudman. 1988. Polls and Surveys. San Francisco: Jossey-Bass.
Cote, Joseph A.. James McCullough, and Michael Reilley. 1985. Effects of Unexpected Situations on Behavioral-Intention Differences: A Garbology Analysis. Journal of Consumer Research, 12 (September): 188-94.
Green, Paul E., Donald S. Tull, and Gerald Albaum. 1988. Research for Marketing Decisions. Englewood Cliffs, NJ: Prentice Hall.
Greenacre, Michael J. 1988. Correspondence Analysis of Categorical Data by Weighted Least Squares. Biometrica, 75, 457-67.
Greenacre, Michael J. 1989. The Caroll-Green-Schaffer Scaling in Correspondence Analysis. A Theoretical and Empirical Appraisal. Journal of Marketing Research, 26, 358-65.
Greenacre, Michael J., and T. Hastie. 1987. The Geometric Interpretation of Correspondence Analysis. Journal of the American Statistical Association, 82, 437-47.
Karande, Kiran W. and V. Kumar. 1995. The Effect of Brand Characteristics and Retailer Policies on Response to Retail Price Promotions: Implications for Retailers. Journal of Retailing. 71 (3): 249-278.
Kellehear, Allan. 1993. The Unobtrusive Researcher - A Guide to Methods. St. Leonards, Australia: Allen and Unwin.
Kumar. V. and Robert P. Leone. 1988. Measuring the Effect of Retail Store Promotions on Brand and Store Substitution. Journal of Marketing Research. 25 (5): May 178-85.
Malhotra, Naresh K, Mark Peterson, and Susan Bardi Kleiser. 1999. Marketing Research: A State-of-the-Art Review and Directions for the Twenty-First Century. Journal of the Academy of Marketing Science, 27 (2), 160-83.
Mayhew, Glenn E. and Russell S. Winer. 1992. An Empirical Analysis of Internal and External Reference Prices Using Scanner Data. Journal of Consumer Research 19, June 62-70.
Noelle-Neumann, Elisabeth and Thomas Petersen. 1996. Alle, nicht jeder Einfnhrung in die Methoden der Demoskopie. Mnnchen: DTV.
Ray, Michael L. 1993. Unobtrusive Marketing Research Methods. Columbus, OH: Franklin University Library.
Ritenbaugh, C. K. and G. G. Harrison. 1984. Reactivity of Garbage Analysis. American Behavioral Scientist, 28 (September/October): 51-70.
Shimp, Terence A. and Alican Kavas. 1984. The Theory of Reasoned Action Applied to Coupon Usage. Journal of Consumer Research. 11: December 795-809.
Sudman, Seymour and Norman M. Bradburn. 1983. Asking Questions. SF: Jossey Bass.
Walters, Rockney G. 1988. Retail Promotions and Retail Store Performance: A Test of Some Key Hypotheses. Journal of Retailing. 64 (2): 153-80.
Walters, Rockney G. and Scott B. MacKenzie. 1988. A Structural Equations Analysis of the Impact of Price Promotions on Store Performance. Journal of Marketing Research. 25: Febr. 51-63.
Walters, Rockney G.1991. Assessing the Impact of Retail Price Promotions on Product Substitution, Complementary Purchase, and Interstore Sales Displacement. Journal of Mareting. 55 (April), 17-28.
Webb, Eugene J., Donald T. Cambell, Richard D. Schwartz, and Lee Sechrest. 1966. Unobtrusive Measures - Nonreactive Research in the Social Sciences. Chicago: Rand McNally.
Marcus Schmidt, University of Southern Denmark, Denmark
Niels Krause, The Danish Commissioner of Police, Denmark
E - European Advances in Consumer Research Volume 5 | 2001
Exploring the Intersection of Digital Virtual Consumption and Family Rituals
Linda Tuncay Zayer, Loyola University Chicago, USA
Jenna Drenten, Loyola University Chicago, USA
C5. Krabby Patties, Kelp Chips, or KitKats?: Exploring the Depictions of Food Featured in Children’s Television Shows
Kathy Tian, University of Illinois at Urbana-Champaign, USA
Regina Ahn, University of Illinois at Urbana-Champaign, USA
Michelle Renee Nelson, University of Illinois at Urbana-Champaign, USA
How Residential Mobility Influences Donations
Yajin Wang, University of Maryland, USA
Amna Kirmani, University of Maryland, USA
Xiaolin Li, University of Texas at Dallas, USA
Nicole Kim, University of Maryland, USA