Usage - Situational Influences on Perceptions of Product Markets: Response Homogeneity and Its Implications For Consumer Research

ABSTRACT - Results of an empirical investigation in the financial services market illustrates a relatively high level of homogeneity of customer preferences when the usage-situation is held constant. This suggests that small sample sizes may be adequate for consumer researches adopting situational control (obtaining customer judgments when the situation in which products/services are to be used is specified.)

Citation:

Rajendra K. Srivastava (1980) ,"Usage - Situational Influences on Perceptions of Product Markets: Response Homogeneity and Its Implications For Consumer Research", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 644-649.

Advances in Consumer Research Volume 7, 1980     Pages 644-649

USAGE - SITUATIONAL INFLUENCES ON PERCEPTIONS OF PRODUCT MARKETS: RESPONSE HOMOGENEITY AND ITS IMPLICATIONS FOR CONSUMER RESEARCH

Rajendra K. Srivastava, University of Texas at Austin

[The author appreciates the helpful comments made by Allan D. Shocker (University of Pittsburgh), Richard Staelin (Carnegie - Mellon University) and Robert Stack (Equibank, N. S.) in the formation of a larger study (which this paper is part of), and the financial support furnished by Equibank, N. A., Pittsburgh. Because of the proprietary nature of the data, some of the competitive relationships between services are disguised.]

ABSTRACT -

Results of an empirical investigation in the financial services market illustrates a relatively high level of homogeneity of customer preferences when the usage-situation is held constant. This suggests that small sample sizes may be adequate for consumer researches adopting situational control (obtaining customer judgments when the situation in which products/services are to be used is specified.)

INTRODUCTION

The understanding of customer choice processes and the competitive relationships in the market place has been the focus of considerable consumer and marketing research. Most of this research is based on the paradigm that choice is a function of product attributes (costs/benefits) and customer characteristics. While perspectives based on products and customers are useful, they appear not to be self-sufficient. Products and customers do not exist in a vacuum. Both are embedded in an environment. Products may be viewed as fulfilling customer requirements/goals that may have been instigated by factors that are internal (such as needs, drives) and/or external (such as societal norms, social influence processes) to the individual.

As early as 1934, situational or environmental factors were recognized as important mediating variables which explained some of the discrepancy between attitudes and action (La Pierre 1934). In recent years, researchers in both the psychological tradition (Fredriksen 1972, Wicker 1972) and consumer behavior (Belk 1975 and 1979, Lutz and Kakkar 1976, Bearden and Woodside 1976, Srivastava et al 1978) have further illustrated the importance of situational variables. However, much remains to be done toward drawing implications for consumer and marketing research.

There are two prime problems hindering systematic research of environmental variables and their effect on customer choices: (1) the lack of a generally acceptable taxonomy of situations, and (2) the large amount of data required from respondents (a large number of judgments are required under different situational contexts). The latter factor results in a much higher cost of data collection per respondent than when situational controls are not instituted.

The concept of a general taxonomy of situational influences, for the moment, appears to be a little farfetched. This is because situational influences that affect some consumer behaviors (e.g., the choice of "high-involvement" products such as automobiles) could be entirely different from those that affect other behaviors (e.g., the choice of "low-involvement" products such as soft drinks)(Belk 1979). However, product-specific taxonomies may be easily constructed by following procedures outlined in Fredriksen's article (1972) and as implemented by Srivastava et al in the "breath freshener" market (1978) and Belk in the "clothing" market (1979).

These efforts in developing product-specific situational taxonomies have resulted in encouraging results. First, it appears that generally two or three dimensions or taxonomic factors are adequate for explaining most of the situational variance. Second, these factors are relatively stable across individuals, and individuals appear to have similar response patterns. It is the intent of this paper to examine the extent of, and the implications due to, response homogeneity across subjects. If these results are consistent in other product markets, it would mean that data collection efforts would be relatively manageable. Parsimonious product-specific taxonomies would allow the generation of a small number of situational scenarios (to represent taxonomic cells) which would capture a large proportion of the situational variance. More importantly, the relative similarity (homogeneity) of response patterns across individuals should allow the use of smaller sample sizes.

Environmental Variables and Response Homogeneity

Consumer researchers have generally ignored situational influences in obtaining "overall" attitude/preference measures. The discrepancy between preference and choice of individuals has been generally classified as "error variance", while the differences in preferences across individuals has been attributed to the "heterogeneity" of customer requirements, needs, goals. Some of error variance and heterogeneity may be due to the fact that choices are situation-specific (while the elicited preferences were not obtained as such), and different individuals may have given their preferences with varying usage or consumption situations in mind. For example, if a respondent is provided several brands of instant and regular coffee and asked to rank order preferences, she/he may do so keeping "flavor" in mind and provide higher ranks to the regular coffee brands. Another respondent may perform the task with "ease of preparation" in mind and provide relatively higher ranks to the instant brands. However, both may actually use instant coffee when in a hurry and regular coffee while entertaining. The above self-serving and somewhat simplistic example illustrates that both error variance and respondent heterogeneity may be attributable the ambiguity of the response task. Accordingly, as argued by Rokeach and Kliejunas, we must elicit attitudes toward objects within a situation if we are interested in predicting situation specific behaviors (1972).

The specification of the situational variables may be expected to more clearly define the attributes that customers seek, and further, simplify the judgmental task. Consequently one may expect greater homogeneity between respondents and less error variance, That is, the explicit consideration of situational influences should allow for a more deterministic view of customer behavior.

If one is to develop an understanding of the effect of situational influences on customer choice on the basis of judgments provided by customers (i.e., by survey methodology rather than experimentation), it is important to ensure that the situations under which responses are elicited are relevant (realistic; occur routinely or systematically) and that respondents have the ability to provide valid responses (i.e., have the information requested). That is, we should be primarily interested in tapping a customer's past experience regarding the use of products in those types of situations that can be anticipated by him/her. For example, Hansen distinguishes among (point-of-) purchase, communication and consumption (or usage) situations (1976). Since point of purchase (promotional discounts, displays) and communication (advertising, publicity) situations and their effects cannot be typically anticipated by customers (Belk 1979), this study restricts its attention to consumption or usage situations (Srivastava et al 1978, Belk 1979).

Objectives

The main purpose of this study is to examine the "adequacy" of small sample sizes (and the relative homogeneity of customer responses) when situation-specific preferences are obtained. In this study the setting is consumer perceptions of product-market structures.

The composition of competitive product-markets is of considerable practical relevance for a number of managerial and public policy questions (Day and Shocker 1976). Srivastava et al (1978) developed a procedure for the construction of product-market structures (as perceived by consumers) where they: (1) develop a product-specific situational taxonomy, (2) develop situational scenarios to correspond to morphological combinations of the taxonomic cells, and (3) obtain measures for the appropriateness of products in usage-situations (scenarios). The last stage furnishes a product by usage-situation matrix which is analyzed to yield competitive product market structures.

The examination of the response patterns (product-usage associations) should enable us to address the main purpose of this study, i.e., the "adequacy" of small samples and the relatively high homogeneity of responses when situation specific preferences are obtained. Additionally, since competitive products are more or less appropriate for use in similar situations, a relatively high response homogeneity across respondents should result in stable product market structures for varying sample sizes -- the corollary purpose of this paper.

The "adequacy" of small sample sizes is interpreted in the light of the amount of error variance that a researcher is willing to tolerate. It is expected that product market structures will asymptotically converge towards some "true" structure as sample size increases beyond a relatively small number (n=40) as the "marginal contribution" of each additional respondent toward aggregated data will decrease as sample size increases.

METHODOLOGY AND RESULTS

The methodology progressed through two stages. The purpose of the first stage was the determination of a usage-situational taxonomy and the subsequent generation of situational descriptions (scenarios corresponding to the taxonomic cells). The second stage involved collection of product-usage associations for a relatively large sample size. The product-usage associations were then averaged across individuals for (a) the full sample, and (b) subsamples (drawn randomly, with replacement) of various sizes. If the response patterns are relatively homogeneous, the product market structures obtained by analysis of the averaged product-usage matrices should rapidly converge to the "true" product market structure (for the full data) as sample size increases. The degree to which the market-structures converge can be gauged by determining "coefficients of congruence" (Harman 1967, Cattell et al 1966).

The "financial services" market was chosen for this study because financial services tend to be "complex" and "intangible". They thus provide scope for greater variability in customer perceptions (Sasser 1976) and, consequently, a stiff test for the proposition that small sample sizes are adequate if situation-specific preferences are elicited. Additionally, though scattered pieces of literature have looked at competition between types of installment loans and credit cards, little is known about the extent of competition among financial service categories (e.g., installment loans versus credit cards). The comprehensive market-structures provided by this study should represent an important step toward organization of literature in the turbulent financial services market.

STAGE 1. Generation of a Usage Situational Taxonomy

[This section is described rather briefly due to space limitations. Details of the procedure (when utilized in the breath freshener market) may be found in Srivastava, Shocker and Day (1978).]

Using the iterative procedure developed by Stefflre (1971), individual and group interviews were conducted to generate initial lists of services and usage-situations. Interviewees were given "bank credit cards" as a "target" service and were asked to suggest as many uses as possible. They were then asked to suggest additional services appropriate to these same uses, additional uses for the expanded service list and so on. In addition, information was gathered regarding circumstances under which service "A" would be used but not "B" (and vice versa). This provided insights regarding the more functional differences between the service alternatives. The above insights and situation and service lists were recorded by respondents on a semi-structured questionnaire.

The iterative procedure resulted in long lists of usage-situations and services. The services alternatives represented various methods that could be used for making payments and ranged all the way from ready cash or cash-on-hand to credit cards, savings accounts, and second mortgage loans. These lists were reduced using the insights developed above as well as insights furnished by bank managers. For example, the researchers' understanding of customer perceptions implied that checking accounts from different banks were functionally equivalent and would be used in the same situations. This resulted in a set of 27 services and 27 usage-situations. Examples of usage-situations and services are provided in exhibits 1 and 2, respectively.

EXHIBIT 1

EXAMPLES OF USAGE -- SITUATIONS

EXHIBIT 2

EXAMPLES OF FINANCIAL SERVICES

A small (n = 30) non-student convenience sample was chosen and data was collected via personal interviews to obtain ratings of services (i) for use in each of the usage-situations (j) along a dichotomous (appropriate/ inappropriate) scale. The proportion of respondents indicating a service as appropriate for a given situation was adopted as a measure of the "degree of appropriateness" (a_ij, as formulated in Srivastava et al 1978).

Taxonomic Approach.  In order to develop a taxonomy of usage-situational influences, the products by usage-situations matrix (cell entries a_ij reflecting the degree of appropriateness) was analyzed by the method of principal components (eigenvalues 13.86, 8.43, and 2.37) explained about 90 percent of the total variance. Each situational description was "content analyzed" and coded in terms of "objective" variables representing the amount of money to be paid, where the situation occurred (local vs. not local; retail vs. non-retail), the amount of time available to make payments, whether the expense was anticipated, whether an additional source of money could be expected to meet the expense...etc. The factor loadings of situations were then correlated with objective codings of situations in order to interpret the reduced situational space. This analysis revealed that there were primarily three situational characteristics that explained most of the variance in the appropriateness of services. These were:

1. the dollar amount required as payment (three ranges were relevant: $10 - $150; $400 - $600; $1,000 -$1,500).

2. the location (local vs. out of town or not local)

3. retail vs. non-retail settings ("retail" being defined as those establishments that can be expected to have their own credit departments).

The three dimensions result in (3 x 2 x 2 = 12) twelve taxonomic cells. New situational descriptions were generated to correspond to these taxonomic cells. In order to check whether these descriptions correspond to their respective cells a group of six respondents were given the descriptions along with the typology (above) and asked to code each description back into the typology. These was substantial agreement and only minor adjustments in the situational descriptions were necessary.

Of the 27 services, 24 were retained for the second stage. The three that were deleted were rarely rated as appropriate for the usage-situations.

The purpose of this stage was to develop a usage-situational taxonomy for the financial services market. This usage-situational taxonomy enables the systematic investigation of the "adequacy" of small sample sizes for research efforts eliciting usage-situational specific preferences.

STAGE 2. Adequacy of Small Sample Sizes

Data Collection and Sampling. The sample for this stage was chosen from a consumer panel listing in a large Eastern city in May 1978. While panel members are not "average customers" due to their greater knowledge- ability, exposure to product tests...etc., they can be expected to fill out more sophisticated questionnaires. Further, data collection costs would be lower because a combination of mail survey and telephone interview modes could be employed (panel members could be expected to follow telephone instructions more rapidly). Finally, the panel listings provided pre-recorded demographic information. This was a relevant consideration as it was expected that the judgmental tasks would require (5 minutes per scenario x 12 scenarios = 60 minutes) approximately an hour to complete and it was considered unwise to further tax respondents. Since a high degree of response homogeneity was expected on a prior basis, the sampling frame was selected on grounds favoring costs rather than representativeness of the population.

The managerially relevant audience (for the "target" service, i.e., bank credit cards) necessarily corresponded to higher socio-economic groups. Accordingly, the following screening criteria were used to select a subsample from the panel listings: (1) education of household head -- at least high school graduate, (2) household income -- at least $12,000 per annum, and (3) age of household head -- between 25 and 50 years. 200 questionnaires were mailed of which 131 were returned after two follow-up calls. This response rate is somewhat low for panels but can be attributable to: (1) an unusually lengthy survey instrument, and (2) the fact that the questionnaire was to be filled by the person responsible for most of the major (approximately over $500) financial decisions for the household (since this would usually require the male in the household to be involved in the response task, a lower response rate can be expected).

While space limitations do not allow a detailed discussion of non-response analyses, the panel records showed that non-respondent families: (1) were likely to have lived in the area for shorter time periods, and (2) had a greater number of bank credit and a smaller number of retail credit cards. These differences appear to indicate that non-respondents tended to be more "mobile" and consequently had fewer local affiliations (retail credit cards tend to be centered "locally", except for a few national chains). There were no significant differences (at the a = 0.10 level, or better) between respondent and non-respondents in terms of demographics such as income, education, age, occupation of the household head, family size and the number of savings and checking accounts.

Generation of Product Market Structures. Of the 131 questionnaires that were returned, 115 were complete and usable. Responses were aggregated across individuals to obtain a service by usage-situation matrix of size (24 x 12). As in the previous stage the cells of this matrix (A) have entries (a_ij) equal to the proportion of respondents who rated a service i as appropriate in situation j. Additional (such) matrices were obtained by random selection, with replacement, of 10, 20,...80, 90 percent subsamples. These matrices will henceforth be referred to as A¹⁰...A^XX...A⁹⁰. Finally, another set of matrices were obtained by adding "error" to the "true" matrix ("true" being defined as the full sample data matrix). [This limiting sample size of 115 was considered adequate due to the relative homogeneity of the responses.] If o2A = 0.043) is the variance of the elements of the true matrix, error components of variance o2E were added at random to the elements a_ij to generate a new matrix of elements according to the relationships:

        b_ij  =   a_ij   +  e_ij

and o²_B  =  o²_" + o²_E

By choosing o2E to be 1, 5, 10, yy...50 percent of o2A, the matrices B¹, B⁵, B¹⁰...B^yy...B⁵⁰ can be easily generated.

The matrices (A, A^xx, and B^yy, above) were reduced by the method of principal components with situations as variables, and factor scores corresponding to services retained for components with eigenvalues greater than unity (this resulted in the retention of three components together explaining approximately 90 percent of the variance in each case). It was argued earlier that if response patterns are relatively homogeneous across individuals, the product market structures (location of points representing services in situational space --i.e., a three dimensional plot of the factor scores) [For arguments regarding the use of principal components analysis and the subsequent plotting of factor scores, see Srivastava et al 1978.] should rapidly converge to the "true" structure as sample size increases.

Convergence of Product Market Structures.  Multidimensional scaling techniques (such as principal components analysis) generally result in an arbitrary choice of axes due to rotational procedures used to approximate "simple structure." This lack of orientation (or correspondence of axes) between two structures may result in the conclusion that the two perceptual spaces are dissimilar when, in fact, they may be quite the same. Cliff (1966) illustrates the above argument and has suggested a least squares procedure for rotating a "data" metric to maximal congruence with a "target" matrix. [This procedure has been operationalized in a computer algorithm: CMATCH. See Carmone (1966) for details.]  It should be adopted before computing "congruence coefficients.''

The congruence co-efficient (between two matrices) suggested by Harman (1967) takes the form:

EQUATION

when X and Y are the two matrices. This co-efficient can range between +1 (perfect agreement) through 0 (no agreement) to -1 (perfect inverse agreement). According to Percy (1976) this co-efficient focus on the directional aspects of the spaces (matrices) and not the levels. In order to account for the similarity of levels it is possible to compute pairwise distances between points in perceptual spaces, and subsequently a correlation co-efficient (R_d) between corresponding pair-wise distances in different spatial representations. R_c has ill-defined distribution characteristics while R_d is distributed as Chi-Square. Though R_d can be used to gauge the statistical change of two spatial representations being different, it still does not provide a "substantial sense" of the level of agreement or disagreement between the two representations. It is for this reason that the matrices B^yy were generated. Thus if a researcher is willing to tolerate only 10 percent error in the data in a "substantial sense", [NOTE: this is not the same as the "significance level" associated with R_d which refers to the probability that the two structures are significantly different in a statistical sense.] s/he could obtain the values of R*c and R*d between factor scores from A and B¹⁰, the researcher willing to tolerate 10 percent error should find the sample proportion (xx) acceptable.

The average congruence coefficients (R_c and R_d) obtained by comparing factor scores resulting from A with the factor scores from A^xx and B^yy are plotted in Exhibit 3. These "averages" are somewhat limiting as only three runs were made for each sample proportion (xx) and error proportion (yy). They will, however, suffice for illustrative purposes -- especially since the values of Rc and Rd that were obtained were relatively stable.

As illustrated in Exhibit 3, a researcher willing to tolerate 10 percent error in the data (R*c = 0.98, R*d = 0.987) should be willing to accept a sample proportion (xx) over 40 percent (i.e., a sample size of 115 x 0.40= 46).

Discussion of Congruence Coefficients for Product Market Structures. The consistently high and gradually converging towards unity values of congruence coefficients (Rc increased from 0.948 to 0.998 as the sample proportion increased from 10 to 90 percent; Rd exhibited a corresponding increase from 0.939 to 0.999) may have several explanations:

(1) since the elements of the product usage matrix (A) are based on measures of concurrence, the marginal contribution of each additional respondent towards determining the proportion of respondents who considered a service appropriate for a given situation (degree of appropriateness) is inversely proportional to the sample size. That is, when the sample size is 10, the 11th respondent has ten times the impact on determining the degree of appropriateness compared to the 101st respond-end (when added to a sample of 100). This may explain the rapid convergence to the "plateau" in Exhibit 2 for curves corresponding to the congruence of A and A^xx.

(2) the relatively high values (for even small sample proportions) may be due to a high degree of homogeneity of responses across individuals. There is reason to believe that this may be the case because when the (12 x 24 = 288) responses of individuals were "strung out", and the individuals "factored", a single dominant factor emerged -- indicative of respondent homogeneity.

EXHIBIT 3

PLOT OF CONGRUENCE COEFFICIENTS

(3) the method of analysis (principal components) may be insensitive to minor fluctuations of cell values. This argument is supported by Percy (1976) who found that when a five-point scale was collapsed into a dichotomous one, the factor solutions obtained were almost identical.

Face Validity of Results.  The factor scores corresponding to services obtained from the full data matrix, and a 50 percent subsample matrix are plotted for the first two dimensions in Exhibit 4. The former are represented by "dots" (.) and the latter by "pluses" (+). The two are encircled and a brief service description label typed alongside. In addition, correlations between the taxonomic dimensions (retail vs. non-retail; local vs. not-local; dollar amount) and factor loadings are indicated by dotted arrows. The length of the arrows depict the "size" of the correlations. Note that the taxonomic dimensions appear not to be orthogonal because they are embedded in a three dimensional space, and are not perfectly aligned with the derived principal components.

The close proximity of the corresponding "dots" (.) and "pluses" (+) provides a physical "interpretation" of the meaning of high congruence coefficients (R_c = 0.991; R_d = 0.993 between A and A⁵⁰). Exhibit 4 appears to be rather complex but is quite easily interpretable. The large encircled dot in the upper left hand quadrant represents a usage-situation characterized by a high dollar payment in a "local" and "non-retail" setting (plotted by virtue of its factor loadings.) The appropriateness of services for this situation may be approximated from this map by projecting the services on a line drawn from the origin through the situation. The distance from the origin to the project then gives a relative estimate of appropriateness. Thus for the situation in question, the mapping appears to indicate that two savings account modes are most appropriate followed by the use of a checking account. This was confirmed by examining the actual measures for the degree of appropriateness. Other situations have not been positioned on the map to prevent further cluttering.

EXHIBIT 4

PLOT OF SERVICES IN SITUATIONAL SPACE

The "matching" of situational characteristics and service attributes lend face validity to the market structure. For example, retail installment loans and credit cards appear to be best suited for "retail" purchases --followed closely by bank and travel and entertainment credit cards. Similarly, bank credit cards and traveler's checks are better suited for out-of-town (not local) usage situations. The multidimensional market structures can aid strategic decisions by providing:

(1) a comprehensive understanding of the market place. Services that are more substitutable will appear closer together in space. Though some of the results were highly intuitive, it is interesting to note that retail installment loans are more competitive with bank and retail credit cards rather than with other types of installment loans. The result was not "obvious" to bank management personnel, though in retrospect they believed it made sense as retailers had the inherent advantage of temporal and distance proximity in the customer's choice process, as well as a greater willingness to finance such purchases.

(2) an interpretation of a service's current position. The strength/weakness of services may be examined by gauging its appropriateness for different types of usage situations. For example, bank credit cards were considered less appropriate in local (as opposed to out-of-town) situations as indicated by its position in Exhibit 4 (and confirmed by examination of the relevant a_ij's). This somewhat surprising result may be an outcome of the advertising emphasis on out-of-town scenarios for use of bank credit cards. Should banks be interested in reducing check usage with an offsetting increase in credit card usage, a change in advertising emphasis would appear to be necessary.

(3) directions for reduction in cannibalization. The cluster of services in the upper right hand quadrant corresponding to convenience credit modes (cash advance on bank cards, check credit, overdraft protection) are often kept as separate accounts by banks. An integration of these services into a single account would not only allow a more effective use of advertising dollars (advertising each separately would encourage cannibalization), but will also simplify account keeping and data processing at the "production end."

Space limitations preclude a more detailed discussion of managerial implications. Also, it must be pointed out that the representation of services in situational space merely "reduces" the original data to a more comprehensive form. This reduction necessarily leads to a loss of data. Consequently, the use of the original data (matrices) is encouraged once the tentative implications are drawn from the mappings.

IMPLICATIONS FOR CONSUMER RESEARCH

This research has shown that when the effects of usage situations are controlled for, the extent to which services are deemed appropriate for use is relatively stable across respondents. This results in a high degree of homogeneity among respondents, and, consequently, the high congruence between product market structures based on small sample sizes. It would seem that the larger amount of data required per respondent is offset by the "adequacy" of smaller sample sizes. Though this research does not test any "hypothesis" regarding what is an adequate sample size, the readers, by using the plots in Exhibit 3, may decide for themselves what the adequate sample size would be corresponding to their own tolerance for error.

Since the same taxonomic procedure that was applied to construct product market structures can be applied in developing usage-situational typologies, such typologies can be constructed using relatively small sample sizes. In addition, the small number of dimensions (n = 2 or 3) that have emerged in this study and others indicates that usage situational influences may be quite easily incorporated in consumer research.

The plots provided in Exhibit 3 should not be taken literally across all research situations because the distributions of the congruence coefficients (R_c and R_d) for different levels of "error" are dependent on: (a) the number of rows and columns of the original matrices (in this case 24 and 12, respectively), (b) the number of factors retained (in this case 3), and (c) the number of scale positions for the preferences obtained. A detailed discussion of the concepts involved is provided by Staelin and Gleason (1972).

While usage-situations appear to dictate the benefits desired, or the relative importance of such benefits, these effects admittedly appear at the product variant/type level of the product hierarchy (see Lunn 1972). This research did not focus at the brand level differences due to the large number of service alternatives that would have to be included. Though the same procedure could have been repeated within submarkets to examine brand level differences, it is questionable whether the same degree of homogeneity would have been observed since: (1) individual differences in brand perceptions are likely to be more pronounced and (2) usage-situations are potentially less likely to effect the choice among alternatives that are more or less similar in terms of their functional capabilities.

The apparent homogeneity of customer responses suggests the possibility that differences in the situations that customers face, rather than differences among the customers themselves, may be better explanators of the differences in choice among customers. Consequently, the usage-situation rather than customer characteristics may be a more stable basic unit of analysis in segmentation research. Subsequent identification of lifestyle, cultural, demographic and other factors which effect the frequency of occurrence of situations can then be used to target marketing efforts to those people who experience relevant (to product/service in question) situations with greater frequency.

This research has perhaps raised more questions than it has answered. The author hopes that it will stimulate and provide direction for further research in the exploration of situational influences in consumer research.

REFERENCES

Bearden, W. O. and A. G. Woodside (1976), "Interactions of Consumptions Situations and Brand Attitudes," Journal of Applied Psychology, 61, 6, 764-769.

Belk, R. W. (1974), "An exploratory Assessment of Situational Effects in Buyers Behavior," Journal of Marketing Research, 11, 156-163.

Belk, R. W. (1975), "Situational Variables and Consumer Behavior," Journal of Consumer Research, 2, 157-164.

Belk, R. W. (1979), "A Free-Response Approach to Developing Product-Specific Consumption Situation Taxonomies," Proceedings of the AMA/Pitt/MSI Conference on Analytical Approaches to Product and Market Planning (Nov. 1977)

Carmone, F. J. (1966), "CMATCH: A Computer Algorithm for Matching Factor Structures." Drexel University, Philadelphia.

Catlell, Raymond C., M. A. Coulter and B. Tsujeka (1966), "The Taxonometric Recognition of Types and Functional Emergents," in R. B. Catlell (Ed.) Handbook of Multivariate Experimental Psychology, Chicago: Rand McNally

Cliff, N. (1966), "Orthogonal Rotation to Congruence," Psychometrika, 31, 1, 33-42.

Day, George S. and Allan D. Shocker (1976), "Identifying Competitive Product-Market Boundaries Strategic and Analytic Issues," Report No. 76-112, Cambridge, Mass.: Marketing Science Institute.

Frederiksen, N. (1972), "Toward a Taxonomy of Situations,'' American Psychologist, 114-123.

Hansen, F. (1976), "Psychological Theories of Consumer Choice," Journal of Consumer Research, 3, 117-142.

Harman, H. (1967), Modern Factor Analysis, Chicago: University of Chicago Press.

La Pierre, R. T. (1934), "Attitudes vs. Actions," Social Forces, 13, 230-237.

Lunn, Tony (1972), "Segmenting and Constructing Markets," in R. M. Worcester (ed.), Consumer Market Research Handbook, Maidenhead, Berkshire: McGraw-Hill.

Lutz, Richard J. and Pradeep Kakkar (1976), "Situational Influence in Interpersonal Persuasion," in Beverlee B. Anderson (ed.), Advances in Consumer Research, Vol. III, Proceedings of the 1975 Conference, Association for Consumer Research, 370-378.

Percy, L. (1976), "An Argument in Support of Ordinary Factor Analysis," in Beverlee B. Anderson (ed.), Advances in Consumer Research, Vol. III, Proceedings of the 1975 Conference, Association for Consumer Research, 143-148.

Rokeach, M. and P. Kliejunas (1972), "Behavior as a Function of Attitude-Toward-Object and Attitude-Toward- Situation," Journal of Personality and Social Psychology, 22, 194-201.

Sasser, W. E. (1976), "Match Supply and Demand in Service Industries," Harvard Business Review, Nov. - Dec., 1976.

Srivastava, Rajendra, Allan D. Shocker and George S. Day (1978), "An Exploratory Study of Situational Effects on Product Market Definition" in H. Keith Hunt (ed.), Advances in Consumer Research, Vol. V, Proceedings of the 1977 conference of the Association for Consumer Research.

Srivastava, Rajendra (1979), "Situational Influences on Competitive Market Structure: Implications for Marketing Planning," Unpublished Doctoral Dissertation, Graduate School of Business, University of Pittsburgh.

Staelin, R. and Gleason, T. C. (1972), "On the Quality of Principal Components," American Marketing Association Combined Conference Proceedings, Spring and Fall 1972, B. W. Becker and H. Becker (eds.) 34, 484-488.

Stefflre, Volney (1971), New Products and New Enterprises: A report on an Experiment in Applied Social Science, Irvine, Cal: University of California.

Wicker, A. W. (1972) "Processes Which Mediate Behavior-Environment Congruence, "Behavioral Science, 17, 265-277.

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