Classifying Compulsive Consumers: Advances in the Development of a Diagnostic Tool

ABSTRACT - This paper reports on initial efforts to develop, employ and validate a screening measure to identify compulsive buyers. The scale is comprised of 14 psychological and economic measures which best discriminate self identified compulsive buyers from members of the general population. Construct validity for this screening instrument was then assessed by comparing responses from people who were identified as compulsive buyers with a sample of self-identified compulsive buyers on a distinct set of psychological and motivational variables. These variables had previously been found to distinguish compulsive buyers from other consumers. Use of this screening instrument with the general sample of consumers indicates that the incidence of compulsive buying in the population may be as high as 5.9%.

Citation:

Ronald J. Faber and Thomas C. O'Guinn (1989) ,"Classifying Compulsive Consumers: Advances in the Development of a Diagnostic Tool", in NA - Advances in Consumer Research Volume 16, eds. Thomas K. Srull, Provo, UT : Association for Consumer Research, Pages: 738-744.

Advances in Consumer Research Volume 16, 1989      Pages 738-744

CLASSIFYING COMPULSIVE CONSUMERS: ADVANCES IN THE DEVELOPMENT OF A DIAGNOSTIC TOOL

Ronald J. Faber, University of Minnesota

Thomas C. O'Guinn, University of Illinois

ABSTRACT -

This paper reports on initial efforts to develop, employ and validate a screening measure to identify compulsive buyers. The scale is comprised of 14 psychological and economic measures which best discriminate self identified compulsive buyers from members of the general population. Construct validity for this screening instrument was then assessed by comparing responses from people who were identified as compulsive buyers with a sample of self-identified compulsive buyers on a distinct set of psychological and motivational variables. These variables had previously been found to distinguish compulsive buyers from other consumers. Use of this screening instrument with the general sample of consumers indicates that the incidence of compulsive buying in the population may be as high as 5.9%.

INTRODUCTION

Research on the phenomenon of compulsive buying was introduced into the consumer behavior literature by Faber, O'Guinn and Krych (1987) and subsequent work has extended those initial findings (Faber and O'Guinn 1988; O'Guinn and Faber 1988; Valence, d'Astous and Fourtier 1988; d'Astous and Tremblay 1989). This abnormal form of consumer behavior is typified by chronic buying episodes of a somewhat stereotyped fashion in which the consumer feels unable to stop or significantly moderate the behavior. Although compulsive buying may produce some short term positive feelings for the individual, it ultimately is disruptive to normal life functioning and produces significant negative consequences (O'Guinn and Faber 1988).

While efforts have been made to document and understand the problem of compulsive buying, none of these previous works have been able to provided an estimate of the incidence of compulsive buying. One reason for this is that much of this work has relied on self-identified subjects. While self-identified compulsive buyers can provide a rich and valuable source of information about this problem, there are some potential limitations and concerns in relying solely on self-identified respondents.

A major problem with only examining self-identified compulsive buyers is that they are likely to be at a later stage in the development of this problem behavior. Research with other forms of addictive and compulsive behaviors indicate that people generally go through denial stages and feelings that they are somehow immune from negative effects before admitting that they truly have a problem (Marlatt et al. 1988; Prochaska and DiClemente 1986; Salzman 1981). People in these earlier phases may differ from those who are able to admit they have a problem. Feelings and behaviors during these earlier periods can be examined by having people try to retrospectively report them, but such efforts are subject to errors in recall and changes in perspective that are likely to come with time and experience. A related problem is that people who seek help may be different from those who don't (Schacter 1982). By relying solely on self-identified respondents, we are unable to examine these possible differences.

Another problem with using self-identified respondents is that it prohibits us from determining the magnitude of the problem. While some anecdotal evidence indicates that compulsive buying may be more common than generally imaged (Jacoby 1986; Mundis 1986), only the development of a screening instrument would provide us with an opportunity to estimate how widespread this problem might be.

A few self-help groups such as Spender-Menders and Spenders Anonymous have published lists of questions to help people determine if they might be compulsive buyers. However, no efforts have been made to assess these as diagnostic tools. One scale, however, has been developed to measure compulsive buying (Valence, d'Astous and Tremblay 1988). This scale is comprised of 13 items which form three dimensions: a tendency to spend, a reactive aspect and post-purchase guilt. This scale has been administered by d'Astous and Tremblay (1989) to a sample of 190 Canadian consumers. However, they simply report bivariate correlations with the measure among their sample, reflecting their view that compulsive buying is not qualitatively but only quantitatively different from normal buying.

Although d'Astous and Tremblay refer to this behavior as an "extreme case of a generalized urge to buy" (1989; 2), they offer no definitional threshold for where this extremity would begin and in their analysis treat it as if it were simply a linear difference in the population.

O'Guinn and Faber (1988), on the other hand, hold that the constellation of attitudes, value orientations and economic behaviors that comprise compulsive buying represent a phenomenon that is fundamentally different from even high levels of the normal buying urge. Instead, they have shown that compulsive buying shares many similarities with other types of compulsive and addictive behaviors (Faber and O'Guinn 1988; O'Guinn and Faber 1988). The debate between a qualitative versus a quantitative difference is admittedly a difficult point to resolve here just as it has been in other areas of clinical and developmental psychology (cf. - Reese and Overton 1970).

Regardless of whether one views compulsive buying to be qualitatively or only quantitatively different from normal buying, a good measure of this trait is needed to distinguish compulsive buyers from other members of the population. This measure should also be able to provide us with an estimate of how common this problem is in the general population. It is, therefore, the purpose of this paper to report our initial efforts at building an effective screening instrument for identifying compulsive buyers and assessing the incidence of compulsive buying within the general population.

METHOD

A survey of both self identified compulsive buyers and members of the general population was employed in this effort. In the case of the problem consumers, the sample was drawn from among people who identified themselves as compulsive buyers and desired, but had not yet received, help for their problem. The sampling frame was explicitly chosen to minimize the number of people already in therapy. - This was done in order to reduce the possibility of contamination in respondent's answers and clinical measures due to learned responses from the particular type of therapy employed and the acceptance of the therapist's explanation for their pathology. Of course, this approach still has problems of self selection bias, and the possibility that self-identified problem spenders will include people who have spending or debt problems but are not truly compulsive buyers. Still. this approach appears to be the best available prior to the development of a valid screening instrument.

The compulsive sample was obtained from people who had written to a California based self-help group for problem consumers. The director and founder of this organization had appeared on several radio and television talk shows throughout the U.S. and had been the subject of numerous newspaper and magazine articles. In the sis months prior to this study, the organization had received approximately 1400 letters as a result of this publicity. All these letters were read and reviewed by the authors, and any letter that did not indicate that the letter writer had a personal spending problem was eliminated. Thus, people who were simply commenting on the interview or those who were writing because someone they knew had a problem, were removed from the sampling frame. The review of the letters identified about 200 that were not written by people with spending problems of their own. These were eliminated from the sampling frame and the remaining letters were stratified by states. A systematic sample, selecting two out of every three letters, yielded a final sample of 808 people who were sent questionnaires. A total of 386 completed questionnaires were returned, for a response rate of 47.8%.

To serve as a comparison group, a sample of the general population of three Illinois cities was drawn. These cities were chosen to represent a large city (Chicago), a small city (Springfield) and a more rural area (Bloomington-Normal and surrounding rural area). A sample of 800 people was drawn and sent questionnaires. After the first mailing, 129 completed questionnaires were received. A second mailing yielded another 121, and a third another 35 for a total of 285 or a response rate of 35.6%.

The questionnaire covered a wide range of dimensions. One set of items was designed to determine if the problem consumers differed significantly from the general population on a measure of compulsivity. This was assessed by a five item subset of the psychasthenia (obsessive-compulsive syndrome) scale of the MMPI. Also included in the survey were measures of materialism, fantasy, sensation seeking and items assessing feelings towards shopping and the consequences of compulsive buying. Finally, several other items were used to assess perceived and objective aspects of economic well being, debt load, credit behavior, and feelings regarding the process of buying. The majority of the items were assessed via five-point Likert type scales. For some items the scales measured the level of agreement (or disagreement) with a statement. Other questions asked respondents to indicate the frequency ("very often"; "often"; "sometimes"; "rarely" or "never") with which they engaged in various behaviors or experienced different feelings.

Analysis

The logic behind the analysis is fairly straight forward. It has sit major steps:

1) Find the variables which best discriminate members of the two strata (compulsive buyers and members of the general population).

(2) Use those variables to construct a screening measure.

(3) Compare the distributions of the screener within both strata.

(4) Compare the degree of intersection of the two distributions and establish a statistically based threshold score for classifying someone as a compulsive buyer.

(5) Estimate the incidence within the general population strata.

(6) Partially validate the screener by comparing the individuals screened from the general population as being compulsive buyers with those who identified themselves as such on a separate set of more general psychological and economic measures.

As is the case in all attempts to estimate incidence of a psychopathology, we are admittedly operating with several very severe constraints. The estimate is thus inferential in the truest sense, and subject to all the same methodological problems that has historically plagued the estimation of prevalence of psychopathology, particularly given our reliance on personality measures (see Nathan 1988 for a critique.)

Previous research by O'Guinn and Faber (1988) and Faber and O'Guinn (1988) has determined a number of predisposing factors or correlates associated with individuals who identified themselves as compulsive buyers. These factors include both psychological (attitudes and emotional states associated with money and shopping, self esteem, general compulsivity, etc.) and economic (i.e., debt load) characteristics. Our task was to develop the most powerful screener possible, then to determine reasonable population parameters for it within the general population, and finally to validate it by using these other variables shown by previous research to be distinguishing characteristics.

Scale Construction

The first step of this process utilized both survey strata (compulsive and general population). A multiple discriminant analysis was used to determine the set of variables which best discriminated the compulsive strata from the comparison strata. Since our goal was to both develop a screening instrument and validate it using the same subjects, some items needed to be used in scale construction while others were reserved for validation. We chose to use items which directly asked about buying behaviors, attitudes and emotions as potential items for the scale. More general constructs which had previously been found to be associated with compulsive buyers were reserved as validation measures.

A stepwise procedure was used with a set of 32 variables which assessed psychological, motivational -and behavioral aspects of buying. The development of these items emanated from two main sources. One source was the existing literature on other forms of compulsive and addictive behaviors. Several items which represented psychological states and behaviors which are commonly found to be present in several other compulsive or addictive behaviors were specifically designed to be applied here to buying situations. The remainder of the items were developed based on our previous knowledge of compulsive buying. Prior to the development of these items, we had sat in on several group therapy sessions with compulsive buyers, talked with psychiatrists and counselors working in this area and conducted both group and individual interviews with compulsive buyers to gain a greater understanding of this problem. The knowledge we gained through these efforts helped us to develop variables which represented behaviors and feelings which many compulsive buyers seemed to share. A pilot test with a small group of compulsive buyers helped us to further refine these measures (Faber, O'Guinn and Krych 1987).

Each variable in the discriminant analysis had to make a statistically significant (p < .10) contribution on the generalized distance measure (Wilk's lambda) to be entered and retained in the function. This level of statistical significance is advisable in the early stages of the construction of a screening instrument when the most powerful discriminating scale is the primary goal. The resultant function was comprised of 14 variables. Table 1 lists these variables in order of descending incremental discriminating power.

The derived discriminant function correctly classified 320 of the 372 people in the compulsive strata (86.02%) and 234 of the 270 in the comparison strata (86.67%), for a total of 86.29% correctly classified. It should, however, be mentioned that since both the solution and the classification were derived from the same data, the classification is to some extent inflated by idiosyncratic error. A future effort employing a separate sample for validation would therefore be worthwhile. Still, it is unlikely that this will reduce the classification power of the function in a dramatic fashion.

Our screening instrument is simply composed of the unweighted summation of the scores on the fourteen variables selected in the discriminant analysis. These are all five point Likert type scales resulting in a screening scale with a range of 14 to 70. Before proceeding further, we tested this scale within the general population for unidimensionality and internal reliability. When factor analyzed only one viable and interpretable factor emerged. The examination of internal reliability determined that Chronbach's alpha was .83. This is considered an acceptable level especially during the development of a new scale.

Incidence

In order to determine the incidence of those predisposed to compulsive buying within the general population, we had to compare the distribution of our screener within the problem shopper strata and within the general strata. I+examining the extent and nature of these two distributions' overlap, we could make some inferences about probable incidence. The determination of the parameter level is essentially an arbitrary one based on statistical criteria. However, this is not only justifiable in a normative sense, in that this is the very way incidence for psychopathologies have been historically determined, but also in the sense that high scores on the screener attests to the presence of a constellation of factors which are only present when this psychopathology exists. We would thus expect very little intersection of scores from the two strata.

By examining the descriptive statistics of the two distributions offered in Table- 2, one can see that only the two opposite tails of the distributions overlap. In other words, only a very few people in the general population have scores high enough on the screener to overlap with people in the compulsive strata. This, was consistent with our expectations about compulsive buying and is very encouraging.

As can be seen in Table 2, the distribution of the screener within the compulsive strata is more normal, while in the comparison strata variance is much lower and the skew toward low levels of compulsivity is great. This is exactly what would be expected in the construction of a screener for a psychopathology that one believes not to be simply high levels of a trait normally distributed throughout the general population, but a uniquely different phenomenon.

As part of the fourth step in our analysis procedure, it was necessary to determine a reasonable cut-point to be used for classification purposes. Since a primary goal in developing a screening instrument is to be able to use it to identify compulsive buyers within the general population, we made this determination based on the distance from the mean screener score within the general population strata. At approximately 1.5 standard deviations from the mean of the this group (a score of 46) one finds 30 individuals or 11.11% of the strata. At two standard deviations (or a scale value of 42) there are 16 of 270 people or 5.93%.

TABLE 1

VARIABLES DISCRIMINATING BETWEEN COMPULSIVE AND NORMAL CONSUMERS

TABLE 2

DESCRIPTIVE STATISTICS FROM 14 ITEM COMPULSIVE BUYING SCALE

We chose to use the two standard deviation criteria. Two standard deviations is admittedly an arbitrary point, but after careful examination of the two distributions, and their intersection, it seemed like a reasonable and fairly conservative choice. People in this tail of the distribution clearly overlap with those in the compulsive strata. A score of 42 also represented the modal score for the self identified compulsive buyers. Thus, people in the general population scoring 42 or lower (lower scores indicated a greater level of compulsive buying) would seem to score similarly on this measure to self identified compulsive buyers.

According to this definitional criterion, approximately 6% of the general population could reasonably be classified as being at risk or predisposed to compulsive buying. Since it seems to be a progressive disorder (d'Astous and Tremblay 1989; Faber and O'Guinn 1988; O'Guinn and Faber 1988) the estimate could be reasonably lowered or raised depending upon-some stage criterion. It is also unlikely, given the mix of variables within the screener which deal specifically with aspects of shopping that we are simply picking members of the population with so called compulsive personalities. Still, our estimate here should be considered a preliminary one based solely on one sample. Yet, despite this strong cautionary note, we feel that the number is meaningful as an initial indication of the potential prevalence of this psychopathology.

VALIDATION

At this point, we had identified 16 individuals that according to our screening instrument, should be at risk, predisposed or behaving as compulsive buyers. If this were so, then we would also expect them to have other characteristics in common with self identified compulsive buyers. Faber and O'Guinn (1988) and O'Guinn and Faber (1988) had found several key characteristics of compulsive buyers. Compared to other consumers, compulsive buyers have a lower level of self esteem, a higher level of generalized compulsivity, higher materialism, particularly on the envy subscale, and a higher fantasy level. These are deeper underlying personality dimensions and value orientations which go well beyond what is measured in the screening instrument, and can therefore be used to establish construct validity for the scale. Another important distinction between compulsive buyers and normal consumers occurs in the economic consequences of the behavior. This can be seen in their much higher debt load. Thus, size of debt load can also serve as an objective economic criteria to provide further construct validity for the screening instrument.

In order to provide partial validation for the compulsive buying scale, we sought to compare the individuals captured by our screener with those who identified themselves as compulsive buyers on these more general dimensions. In order to have near equal numbers for comparison we randomly sampled 16 individuals from within the compulsive strata. These 16 self identified compulsive buyers were then compared with the 16 individuals identified by our screener. A series of t-tests were used to compare the mean responses of the two groups on those dimensions. These findings can be seen in Table 3.

As can be seen, there were no statistically significant differences between the self identified compulsives and the members of the general population identified by the developed screener on these critical dimensions. This is supportive of the belief that the individuals selected from the general population by the screener are no different than those who identify themselves as problem shoppers. Some additional support for the construct validity of this measure is obtained by considering the mean values of the total general population as reported in Faber and O'Guinn (1988). Here, examination of the means reveals fairly large differences between the screened population and population from which it was drawn. For example, the mean value on the self esteem measure was 14.92 for those identified by the screening instrument -as compared to 18.33 for the general population. Most striking was the percent of household income, after mortgage or rent, which goes to pay for past purchases (40 vs. 22). Likewise, the fantasy score was 6.38 for the screened population and 9.15 for the general sample. General compulsivity was 14.31 for the screened population and 10.84 for the general population. Finally, envy was 6.98.in the general strata and 9.50 in the screened compulsives. So, in terms of the validating items, those screened from the general population have mean scores very much like those in the compulsive strata, and very different than those in the general population. Taken together, this evidence suggests that the screener is being quite successful in distinguishing those individuals who have similar psychological and economic attitudes, value orientations and behaviors as compulsive buyers and thus appears to possess construct validity.

DISCUSSION

There are several obvious limitations to this study that should be addressed. First, the initial sample of compulsive buyers was self selected. It's quite possible that because of this, they are significantly different than those who, for various reason, would not choose to identify themselves or cooperate in a study. Yet, this is almost always a problem with clinical-epidemiological studies. Until we are much further along in our knowledge about the problem of compulsive buying, we have no other option but to build our knowledge and our measures based on those people willing to be identified.

A second potential problem arise from the fact that the comparison sample was from Illinois only. While we have no reason to believe that Illinois is significantly different than the rest of the nation, we simply do not know this. Again, this awaits a larger study. Since our instrument is built on the basis of responses from these two groups it is subject to errors which might arise from any differences between them and the larger populations they are assumed to represent. Furthermore, there may be many other important variables and dimensions yet unexplored which would better screen for a predisposition to compulsive buying. For these reasons we would encourage further efforts to validate our initial attempt at a screening instrument and develop improved versions.

TABLE 3

There are also several important things to be done in future research. One is to more directly confront the tangle that has plagued the study of all addictive behaviors and compulsions in that causality is so terribly illusive. For example, it may be that compulsive buying causes low self esteem, rather than the other way around. Still, causality is not a necessary requisite for building an effective screener, which is our most immediate goal. Most importantly, we need to move more in the direction of examining the psychological, societal, cultural, and environmental processes operative in compulsive buying. For now, a descriptive and phenomenological approach is entirely appropriate, but we must someday move beyond that.

CONCLUSION

We believe consumer researchers should be concerned about the abuse potential of consumption and its relationship to all sorts of factors such as environmental (retail) stimuli, mood states, etc. As part of this concern, it seems reasonable to want to know something of the prevalence of compulsive buying. Just how "deviant" is the behavior? How many are suffering from it? Is it getting more common? Does it exist in other cultures? How concerned should policy makers be? To answer these types of questions we need to have some idea of incidence levels. Our best estimate based on the screener developed here is that as much as 5.9% of the population may be at risk for becoming compulsive buyers. Although this figure may seem high, it is in line with estimates of the incidence of other types of addictive and compulsive behaviors.

While we believe that this estimate represents a significant number of people which further justifies the need to study compulsive buying, we frankly also feel that the problem of compulsive buying should be studied regardless of the number of people it effects. We hold this view for at least two reasons. The first is that by studying the abnormal we often learn a great deal about the "properly functioning" organism. Thus, the study of compulsive buying may provide important insights into factors that also influence more "normal" buyer behavior. Secondly, no matter how rare, the fact that consumption seems to have an abuse potential similar to substances or other activities is not only noteworthy, it has been so totally ignored in the consumer behavior literature that it is conspicuous by its absence. As part of our social responsibility, we need to be aware and concerned about issues such as this. It is hoped that with the further development of effective screening instruments, our ability to examine this phenomenon will be enhanced.

REFERENCES

d'Astous, Alain and Sylvie Tremblay (1988), "The Compulsive Side of "Normal" Consumers: An Empirical Study," paper delivered at the 18th Annual Conference of the European Marketing Academy, April, 1989.

Faber, Ronald J. and Thomas C. O'Guinn (1988), "Compulsive Consumption and Credit Abuse", Journal of Consumer Policy, 11, 97-109.

Faber, Ronald J. and Raymond Krych (1987), "Compulsive Consumption," in Advances in Consumer Research, Vol. XIV, eds. Melanie Wallendorf and Paul Anderson, Provo, Utah: Association for Consumer Research, 132-135.

Jacoby, Susan (1986), "Compulsive Shopping," Glamour, (April), 318-319; 248-351.

Marlatt, G. Alan, John S. Baer, Dennis M. Donovan and Daniel R. Kivlahan (1988), "Addictive Behaviors: Etiology and Treatment," Annual Review of Psychology, 39, 223-252.

Mundis, Jerrold (1986), "A Way Back from Deep Debt," New York Times Magazine (Jan. 5), 22-26.

Nathan, Peter E. (1988), "The Addictive Personality: How Valid? How Useful?" Journal of Consulting and Clinical Psychology, 56, 332-340.

O'Guinn, Thomas C. and Ronald J. Faber (1988), "An Exploration of the World of Compulsive Buying: Correlates, Process and Consequences," working paper, University of Illinois.

Prochaska, J. O. and C. C. DiClemente (1986), 'Toward a Comprehensive Model of Change," in Treating Addictive Behaviors: Processes of Change, eds. William R. Miller and N. Heather, New York: Plenum Press.

Reese, Hayne W. and Willis F. Overton (1970), "Models of Development and Theories of Development," in Life-Span Developmental Psychology, eds. L. R. Goulet and Paul B. Baltes, New York, Academic Press, 116-145.

Salzman, Leon (1981), "Psychodynamics of the Addictions," in Behavior in Excess: An Examination of Volitional Disorders, ed. S. Joseph Mule, New York: Free Press, 338-349.

Schacter, Stanley (1982), "Recidivism and Self-cure of Smoking and Obesity." American Psychologist, 37, 436-444.

Valence, Gilles, Alain d'Astous and Louis Fortier (1988), "Compulsive Buying: Concept and Measurement," in Proceedings of the XVIlth Annual Conference of the European Marketing Academy, eds. K. Blois and S. Parkinson, Bradford, England, University of Bradford Management Center, 601-624.

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