Questioning the Concept of Involvement Defined Product Classes

John L. Lastovicka, Temple University
ABSTRACT - Homogeneity of consumer acquisition behavior is examined within a set of diverse product classes. The degree to which different levels of involvement are related to levels of acquisition behavior is examined.
[ to cite ]:
John L. Lastovicka (1979) ,"Questioning the Concept of Involvement Defined Product Classes", in NA - Advances in Consumer Research Volume 06, eds. William L. Wilkie, Ann Abor, MI : Association for Consumer Research, Pages: 174-179.

Advances in Consumer Research Volume 6, 1979      Pages 174-179

QUESTIONING THE CONCEPT OF INVOLVEMENT DEFINED PRODUCT CLASSES

John L. Lastovicka, Temple University

ABSTRACT -

Homogeneity of consumer acquisition behavior is examined within a set of diverse product classes. The degree to which different levels of involvement are related to levels of acquisition behavior is examined.

INTRODUCTION

Low involvement theorists (e.g., Krugman, 1965; Ray et al, 1973; and Robertson, 1976) suggest that much of the complexity inherent in studying consumer behavior is due to the broad assortment of products acquired and used by consumers. Accordingly, low involvement theory has categorized product classes and suggested different types of purchase behavior in each product category. In one category are the so called "low involvement product classes." Descriptions of typical purchase behavior for this product category have been different from those as described by the traditional consumer behavior models.

Frequently purchased, commodity-like goods such as toothpaste or canned peas are examples of what have become referred to as low involvement product classes. Following Lastovicka and Gardner (1978), a low involvement product class is one in which most consumers perceive little linkage to their important values and is a product class where there is little consumer commitment to the brands. Less frequently purchased, and more brand differentiated product classes, such as automobiles or stereo equipment are often given examples of high involvement product classes. More direct linkage to personal values and more commitment to brands have been assumed in these latter product classes.

Please note that an involvement defined product classification system is consumer and not product defined. That is, a particular good's inclusion in the low involvement group is not determined by any objective characteristics of the good itself - as in Aspinwall's (1961) "color theory'' product classification system. Rather, homogeneous consumer perceptions and behaviors are what determine a particular good's involvement classification.

Low involvement theorists have used the involvement concept as the key variable in explaining differences in behavior across different product classes. In purchasing brands in low involvement product classes, buyer behavior is assumed to be passive, re-active instead of active problem solving. Kassarjian (1978a, 1978b) has suggested that Skinner's (1972) behavioral modification approach may be more than enough to explain brand choice in low involvement product classes. The cognitively based traditional models of brand choice behavior are felt to be most appropriate in the high involvement product classes.

THE RESEARCH QUESTION

The present study is an attempt to question the notion of an involvement based product classification scheme. Two criteria were devised to test the classification system.

Criterion One: Generalizability

A primary concern is of generalizability. In order for any product classification system to be useful, products in a like classification must share similar consumer behavior. Products of different classifications should demonstrate heterogeneity. The degree of homogeneity within a given involvement classification (e.g., the homogeneity of behavior with a product in the low involvement classification) is questionable as the involvement product scheme is consumer defined. Given knowledge of the differences in individual values, perceptions and behaviors it is likely that same product behavior may not be similar across individuals.

That is, it is possible that products may not be the object of very similar behavior from individual to individual. If the same product has different meanings to different consumers, then the power of generalizability has vanished. Only if the same product is the object of approximately the same behavior across consumers, can generalizations be made about the product class.

One way of increasing the power of generalizability of the involvement product classification system is to better specify what is being classified. Following Belk (1975), it is recognized that one way to more clearly specify a stimulus is to consider the object-in-situation stimulus. Accordingly, products-in-consumption situations would seem to be a better point of classification than products alone.

Criterion Two: Identification

In order for involvement classification system to be useful, not only must there be homogeneity of behavior ac-cross consumers within products-in-situations, but there must be a basis for believing that involvement is the key explanatory variable for the assumed differences in behavior. Identification of the key variable in the classification system is imperative. If involvement is not a key variable, then the terms high involvement product class and low involvement product class are misnomers.

This identification problem severely weakens the power of the involvement classification system. This is especially true in the case when variables, other than involvement, can be used to explain phenomena that involvement supposedly explains.

For example, Hupfer and Gardner (1971) have measured a great disparity between automobiles and toothpaste on an involvement scale. But is it primarily involvement which is responsible for the greater degree of thought and consideration expected to go into an automobile purchase relative to a toothpaste purchase? An equally plausible explanation would rely upon learning or familiarity. With more time between automobile purchase, more consideration should go into such a choice decision since consumers are more likely to forget much of their prior make and model evaluations since ,the last choice. Moreover, even if memory decay was minimal, market structure - in terms of the makes and models of automobiles available - may have changed considerably since the last purchase. With the toothpaste purchase, less pre-purchase consideration can be explained by greater reliance on the last purchase experience due to increased recency. For this automobile-toothpaste example, the distinction between a reliance on prior learning and re-learning seems as palatable as a distinction based on involvement. This suggests that at least familiarity, a competing key variable, must be tested against involvement.

RESEARCH DESIGN

Sample

During Spring 1978, a systematic sample (Sudman, 1976) of 334 operating household phone numbers was drawn from the Philadelphia Telephone directory. Using a procedure described by Jolson (1977), the telephone subscribers drawn were phoned and a request was made to mail a questionnaire. Permission to mail the questionnaire was not obtained from 76 individuals. From the 258 remaining, 143 useable completed questionnaires were returned. This 143 represents a 43% response rate from the initial sample of 334 and a 55% response rate from those mailed the questionnaire.

Measurement

Low involvement theorists have drawn a sharp contrast in their descriptions of buyer behavior in low involvement and high involvement product classes. For high involvement goods the buyer is characterized as an active problem solver and an information seeker. In contrast, for low involvement products the buyer is seen as a creature of habitual routine. In Howard and Sheth's (1969) terms, the high involvement buyer engages in extensive problem solving while the low involvement buyer engages in routinized response behavior.

Behavior for seven different current or possible future purchase decisions was measured using the extensive problem solving -- routinized problem solving dichotomy. The Appendix shows how the two types of behavior were operationalized in the mail questionnaire. Consumer H, in the appendix, was described as engaged in extensive problem solving while Consumer M was described as a routinized problem solver. Respondents were asked to indicate to what degree they would be like H or M for each of seven purchases. The degree of extensive-routinized response behavior was measured on the following continuum of acquisition behavior:

DIAGRAM

These responses were coded on a 1-5 scale where "very much like H" was set to 1 and "very much like M" set to 5.

The seven product - in - consumption behavior situations were:

(1) A bottle of wine as a gift for a friend.

(2) A bottle of wine to be consumed by myself or family at home.

(3) An automobile for personal use.

(4) Toothpaste for my personal use at home.

(5) A loaf of bread to be used by myself or family at home.

(6) Lightbulbs for use at home.

(7) A high-fidelity stereo sound system for use at home.

Prior research (Hupfer and Gardner, 1971; Lastovicka and Gardner, 1978; and Belk, 1976) suggest that (1), (3), and (7) are in the high involvement domain while (2), (4), (5) and (6) are in the low involvement domain.

Each respondent's own level of involvement and familiarity for each of the six product classes represented in (1) - (7) was measured with four additional questions. Each of these questions consisted of an item using a 1-5 Disagree-Agree scale. Following Lastovicka and Gardner (1978) involvement is seen as having two components: normative importance and commitment. Normative importance is based on Rokeach's (1975, 1969) value system and the degree to which a product is linked to important values. It was measured with:

"I rate this product as being of the highest importance to me personally."

Commitment is based upon Kiesler's (1971) notion of the bonding or pledging of an individual to a particular choice object. It was measured with this item:

"If my preferred brand or make and model was not available at the store or dealership, it would make little difference to me if I had to choose another brand in this general group of products."

Knowledge about the product class was measured with:

"I think that I could talk about this general group of products for a long time;"

while remembered personal experience with the product class was measured with:

"I can remember having purchased something in this general group of products."

Convergent and divergent validity for the first three items of the four presented above is presented by Lastovicka and Gardner (1978).

Measures were also taken to classify the respondents by sex, marital status, occupation, age and income.

ANALYSIS

Generalizability

Using a repeated measures two factor mixed analysis of variance model (Winer, 1971), the criterion of generalizability was tested. The model examined the influence of individual differences and product-situation combinations on one dependent measure (the response on the 1-5 extensive problem solving -- routinized problem solving scale).

The intent of using analysis of variance (ANOVA) was twofold. First, it was used to construct estimates of the relative source contributions in the ANOVA model to variance in the dependent measure. Of prime concern was the relative contribution, due to products-situations versus individual differences, on acquisition behavior. Second, ANOVA was used to assess the significance of a contrast between the three product-situations assumed in the high involvement realm and the four product-situations assumed in the low involvement realm.

Using ANOVA in the first manner required stipulating the expected mean squares and the solutions for the estimate of variance components (Winer, 1971; Kirk, 1968). The expected mean squares and the estimates of variance components are shown in Table 1. Since repeated measures within each product-situation treatment were not taken, it is impossible to separate the error component, EQUATION, from a potential interaction component, EQUATION. Seeking a simple model, the results presented assume that the interaction component,  EQUATION, is zero.

TABLE 1

E(MS) AND ESTIMATES OF VARIANCE COMPONENTS FOR ANOVA MODEL

TABLE 2

RELATIVE CONTRIBUTIONS OF VARIANCE

The results in Table 2 suggest that the products - in - situations do slightly dominate over individual differences in contributing to explained variance. A reasonable interpretation is that differences in acquisition behavior are due primarily to differences in products and how they are to be used and, secondly, to differences between people. Although a large share of the variance, 20.32%, is accounted for by the product-in-consumption-situation notion, a disturbingly large proportion of the variance, 62.35%, is left unaccounted for. This suggests the presence of an interaction component.

Scheffe's multiple comparison test was used to test the difference in mean levels in acquisition behavior between the three so-called high involvement and the four low involvement products-in-situations. The mean level for the former group, 2.02, was significantly lower than the mean for the latter group, 3.10, (p < .05). This suggests that average claimed acquisition behavior with the group consisting of automobile for personal use, wine for gift and stereo for personal use can be best described as "somewhat like" extensive problem solving behavior. For the group consisting of light bulbs for use at home, bread for use by self or family, toothpaste for personal use, and wine for family or self consumption, a description of "like both" extensive and routinized acquisition behavior is appropriate.

Figure 1 is a visual aid in analyzing homogeneity. Figure 1 presents, for each purchase decision, the percentage of the 143 subjects who checked either the 4 or 5 point on the scale (indicating extensive problem solving) versus the mean value of each purchase decision on the same scale. This reinforces the notion that not only are there overall differences in acquisition behavior between purchase decisions, but that the behavior for each type of decision is relatively homogeneous.

FIGURE 1

EXTENT OF HOMOGENEITY OF ACQUISITION BEHAVIOR FOR SEVEN PRODUCTS-SITUATIONS

Identification

So far, seven different products - in - consumption situations have been demonstrated to exhibit different types of acquisition behavior. Furthermore, within products this acquisition behavior is relatively homogeneous ac-cross people. The degree to which involvement is the key explanatory variable for these behaviors, must still be tested.

The influence of the two components of involvement, the measures of knowledge and prior experience, as well as a set of demographic measures were used in a multiple regression model to explain the dependent variable of the 5 point scale measure of extensive-routinized response behavior. Since the ANOVA suggests that the respondents were relatively homogeneous in behavior within a product class and heterogeneous across products, there was some justification to consider the data base as consisting of 858 (6 products-personal-use situations X 143 respondents) independent data cases. That is each respondent supplied 6 data cases, one for each product - personal use purchase, into the data matrix used in the multiple regression analysis. The multiple regression model used was:

X1 = B2X2 +B3X3 + B4X4 +B5X5 + B6X6 + B7X7 + B8X8 + B9X9 + B10X10 + B11X11 + B12X12 + B13X13 + B14X14 + B15X15 + B16X16 + B17X17 + e   (1)

X1 = degree of extensive-routinized problem solving claimed for a given product for personal use (the higher the value on the 1-5 scale, the more routinized).

X2 = degree of knowledge about the same given product (the higher the value, the more knowledgeable).

X3 = perception of the normative importance of the same given product (the higher the value, the greater the importance).

X4 = purchase experience with the same given product (the higher the value, the greater the experience).

X5 = non-commitment to brands in the same given product class (tl~e higher the value, the less committed)

X6 = dummy variable indicating sex (0 = male, 1 = female).

X7 = dummy variable for single or never married marital state.

X8 = dummy variable for married marital state

X9 = dummy variable for divorced or separated marital state.

X10 = dummy variable for professional occupation.

Xl1 = dummy variable for white collar occupation.

X12 = dummy variable for blue collar occupation.

X13 = dummy variable for housewife occupation.

X14 = dummy variable for student occupation.

X15 = dummy variable for retired.

X16 = age in years.

X17 = income range

(1=$5,000 or less,  2=5,000-10,000,  3=10,000-15,000,  4=15,000-20,000,   5=20,000-25,000,  6=25,000 or more)

As a quick review, the variables on the right side of the equation are expected to be related to the dependent measure as follows:

1. Low Involvement Theory Suggests: a negative relationship between normative importance and routinized behavior; and a positive relationship between non-commitment and routinized behavior.

2. Learning and Familiarity Explanations Suggest: a positive relationship between both knowledge and experience and routinized behavior.

3. Other Explanations of Acquisition Behavior Suggest: a positive relationship between age and routinized-like behavior (Hempel, 1969); and a negative relationship between occupational status and routinized-like behavior (Newman and Staelin, 1972).

4. No particular relationships: for sex, marital status, or income were expected. Since a good deal of acquisition behavior was due to the between people component in the ANOVA, these additional standard measures of individual differences were included in an exploratory manner.

Multiple regression analysis was performed on equation (1) for the 680 data cases of the 858 which contained no missing data. The results of this regression equation are presented in the top half of Table 3. The signs of the beta weights for this equation are generally as expected, except for the very notable reverse of the sign of the knowledge weight. The familiarity hypotheses argues that as knowledge increases, routinized problem solving behavior should increase. The empirical results suggest rather that as knowledge increases routinized problem solving decreases. Another set of unexpected results were the significance of the beta weights for the sex, income and marital state measures. The sex coefficient sign suggests that women are more likely to engage in routinized problem solving while men are more likely to be more careful shopping and engage in extensive problem solving. The income beta weight suggests more routinized behavior for those with higher income. And finally from the net effects of the weights for the marital dummy variables, one may believe that only the widowed state, the fourth marital classification, had influence on acquisition behavior. The beta weight for the retired suggests more routinized behavior for this group.

The interpretation about the effects of widowhood suggests that the effects of the three marital statuses directly modeled in equation (1) - single, married, divorced or separated - are the same. Using a procedure as shown by Rao and Miller (1971, p. 145), this was examined by testing the null hypothesis:

B7 = B8 = B9.

Since the obtained F (3.38 with 2 and 666 df) was less than the critical value, the null hypothesis stands. Consequently, the only two relevant marital states were widowed and not widowed. This was modeled with a new dummy variable, X18, representing widowhood.

The independent variables with significant beta weights in equation (1) were retained and used in a second regression equation of the form:

X1 = B2X2 + B3X3 + B4X4 + B5X5 + B17X17 + B18X18 + e   (2)

The results of this second regression equation are in the lower half of Table 3.

The net effect of the regression results of equation (2) is in general support of using involvement as a partial identifying key in this behaviorally based product classification system. First, the direction of relationships between the two components of involvement, normative importance and commitment, and degree of routinized problem solving behavior are exactly as expected. The expected direction of relationship was not, however, obtained with the measure of knowledge. This weakens the power of this competing explanation. Second, the relative importance of the involvement components as defined by their contribution to explained variance is high. Table 4 shows that the two involvement components accounted for 33.10% of the explained variance. Only 10.98% of the variance is properly explained by familiarity. That is, only experience has the expected sign and it explains only 10.98%.

TABLE 3

RESULTS OF REGRESSION ANALYSES

TABLE 4

CONTRIBUTION OF INDEPENDENT VARIABLES TO EXPLAINED VARIANCE IN REGRESSION EQUATION (2)

IMPLICATIONS

This study suggests that an involvement-based product, really product-in-consumption situation, classification has more than face validity. Across the products used in the research, consumers could generally be classified as more or less active in their acquisition behavior. Further, using correlation-based methods, involvement was shown to be strongly related to acquisition behavior.

The study should also serve as a reminder that consumer acquisition of Low involvement products is done without, the commonly assumed, meticulous examination of available brands. Despite the efforts of marketers to differentiate their brands, the lack of commitment suggests that consumers perceive brands in low involvement classes as near perfect substitutes.

An interesting question arises in considering how consumers arrived at the stage of less active acquisition behavior with low involvement products. The traditional consumer behavior models suggest active consumers directly engaging in firsthand extensive and limited problem solving behavior at earlier points in time. The logic of low involvement theory offers the alternative explanation of a more passive consumer who at the point of first purchase is content to rely upon product information that was indirectly "caught" and not directly sought out. Information could be caught in several ways. This could include spectator-like observation of the prior purchase experience of other consumers as well as that information incidentally picked up from repetitive advertising. Such information "catching" is an alternative to the more commonly assumed information seeking.

This study also calls attention to the need to seriously consider an involvement product classification system on a segment-by-segment basis. The large residual variance in the ANOVA reported in the study suggests regularity of behavior, not so much at total market level, but rather less aggregated level. To the degree that market segment differences in product involvement can be systematically understood, the classification system offers more promise.

An immediate implication is that no one classification device is a very satisfactory device for totally accounting for differences in acquisition behavior. Explained variance is low. If product classification systems are to be of real value they must be multifaceted. For example, combination of a market structure approach to product classification (e.g., Aspinwall, 1961) and a behavioral system, similar to what is examined here would seem to provide a useful point for future research. Since any one theoretical approach is lacking, the low involvement theorists must be aware of different explanations of the same phenomena.

APPENDIX

Two consumers, H and M, live in the same neighborhood in Philadelphia. Both purchase the same products, but they make up their minds about what to buy in very different ways. Depending on what is being purchased, you may be like either Consumer H or Consumer M.

CONSUMER H

When going shopping, Consumer H feels that little knowledge can be drawn from previous shopping. Consumer H feels a lack of knowledge about the different brands or makes and models that are available in the stores. H is uncertain as to what features ought to be used in evaluating the different brands.

As a result, Consumer H actively looks for information about different brands as well as ways of evaluating the brands. Very often Consumer H will go out of the way to get information from trusted friends. Consumer H may also look at advertising for information.

Consumer H actively searches for information about the products on an extensive basis. H takes the time to learn about the products. Consumer H shops and compares the different brands.

CONSUMER M

Consumer M has the feeling of extensive knowledge when making a purchase. M knows what the different brands have to offer and, in fact, has fairly strong likes and dislikes for a few brands.

As a result, Consumer M buys with little, if any, real thought. Consumer M does little searching for information before making a purchase. M typically buys without having made comparisons between alternative brands. Consumer M feels that past purchasing experience can be relied upon.

REFERENCES

Aspinwall, Leo V. Four Marketing theories. Boulder: University of Colorado, 1961, chapters 3-4.

Balk, Russell W. "Situational Variables and Consumer Behavior," Journal of Consumer Research, 2 (December 1975) 157-164.

Belk, Russell W. "It's the thought that counts: a signed digraph analysis of gift giving," Journal of Consumer Research, 2 (December 1976) 155-162.

Hempel, D. J. "Search behavior and information utilization in the home buying process," In P. R. McDonald, ed., Marketing Involvement in Society and the Economy. Chicago: AMA, 1969, 241-249.

Howard, John A. and Jagdish N. Sheth, The Theory of Buyer Behavior. New York: John Wiley, 1969.

Hupfer, Nancy T. and David M. Gardner, "Differential Involvement with Products and Issues: An Exploratory study," In D. M. Gardner, ed., Proceedings, Association for Consumer Research, College Park: ACR, 1971, 262-269.

Jolson, Marvin A. "How to double or triple mail-survey response rates," Journal of Marketing, 41 (October 1977) 78-81.

Kassarjian, Harold H. "Presidential Address," In H. Keith Hunt, ed., Advances in Consumer Research, Vol. 5, Ann Arbor: Association for Consumer Research, 1978.

Kassarjian, Harold H. and Waltraud M. Kassarjian, "Attitudes under low involvement conditions," In John C. Maloney and Bernard Silverman, ads., Attitude Research Plays for High Stakes, Chicago: American Marketing Association, 1978.

Kiesler, C.A. The Psychology of Commitment. New York: Academic, 1971.

Kirk, Roger E. Experimental Design: Procedures for the Social Sciences. Belmont: Wadsworth, 1968.

Krugman, Herbert E. "The impact of television advertising: learning without involvement. Public Opinion Quarterly, 29 (Fall 1965), 349-56.

Lastovicka, John L. and David M. Gardner, "Components of Involvement," In John C. Maloney and Bernard Silverman, eds., Attitude Research Plays for High Stakes, Chicago: American Marketing Association, 1978.

Newman, J. W. and R. Staelin, "Prepurchase information seeking for new cars and major household appliances," Journal of Marketing Research, 9 (August 1972) 249-257.

Rao, P. and R. L. Miller, Applied Econometrics. Belmont: Wadsworth, 1971.

Ray, M. L., A. G. Sawyer, M. L. Rothschild, R. M. Heeler, E. C. Strong and J. B. Reed, "Marketing Communication and the Hierarchy of Effects," In P. Clarke, ed., New Models for Mass Communications Research, Vol. II, Beverly Hills: Sage, 1973.

Robertson, T. S., "Low Commitment Consumer Behavior," Journal of Advertising Research, 16 (April 1976) 19-27.

Rokeach, M. Beliefs, Attitudes, and Values. San Francisco: Jossey-Bass, 1975.

Rokeach, M. "Role of Values in Public Opinion Research," Public Opinion Quarterly, 32 (Winter 1968) 547-60.

Skinner, B.F. Beyond Freedom and Dignity. New York: Knopf, 1972.

Sudman, Seymour, Applied Sampling. New York: Academic, 1976.

Winer, B. J. Statistical Principles in Experimental Design. New York: McGraw Hill, 1971.

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