Problems With Vals in International Marketing Research: an Example From an Application of the Empirical Mirror Technique

ABSTRACT - This paper examines the usefulness of VALS by surveying a cross-section of students from both the U.S. and other countries and by presenting, explaining, and applying a VALS algorithm, which was empirically developed to mirror the proprietary VALS algorithm. The results suggest that the demographically adjusted empirical mirror approach to VALS does a reasonable job of reproducing VALS. categorizations. The problems of studying values cross-culturally and the use of VALS specifically are addressed. An alternative methodology, LOV, is suggested as a preferable means to examine value structures.

Citation:

Sharon E. Beatty, Pamela M. Homer, and Lynn R. Kahle (1988) ,"Problems With Vals in International Marketing Research: an Example From an Application of the Empirical Mirror Technique", in NA - Advances in Consumer Research Volume 15, eds. Micheal J. Houston, Provo, UT : Association for Consumer Research, Pages: 375-380.

Advances in Consumer Research Volume 15, 1988      Pages 375-380

PROBLEMS WITH VALS IN INTERNATIONAL MARKETING RESEARCH: AN EXAMPLE FROM AN APPLICATION OF THE EMPIRICAL MIRROR TECHNIQUE

Sharon E. Beatty, University of Alabama

Pamela M. Homer, University of Texas at Austin

Lynn R. Kahle, University of Oregon

ABSTRACT -

This paper examines the usefulness of VALS by surveying a cross-section of students from both the U.S. and other countries and by presenting, explaining, and applying a VALS algorithm, which was empirically developed to mirror the proprietary VALS algorithm. The results suggest that the demographically adjusted empirical mirror approach to VALS does a reasonable job of reproducing VALS. categorizations. The problems of studying values cross-culturally and the use of VALS specifically are addressed. An alternative methodology, LOV, is suggested as a preferable means to examine value structures.

INTRODUCTION

Values are central to people's lives (Rokeach 1973). Many behaviors are enacted to fulfill values. Values influence attitude formation and the way in which information is processed, for example. Recent marketing research has shown that values have a significant influence on, among other things, television viewing habits, activity preferences, store choice, consumer decision criteria, consumer product choice, reaction to discontinuous innovations, and cigarette smoking (Becker and Conner 1981, Kahle 1985, McQuarrie and Langmeyer 1985, Pitts and Woodside 1984).

Values provide clues about how a society operates because values are also central to society. Some scholars view values as the individual's representation of a society's goals. If one wants to understand a culture, investigation of the values of people in that culture provides a promising starting point. Values should logically be a central topic for cross-cultural research. A recent study (Kahle 1986) demonstrated that values vary among geographic regions in the U.S. Nevertheless, few scholars have published rigorous studies designed to investigate comparative values internationally (Berrien 1966; Zavaloni 1980).

International Research on Values

The opportunities and needs for international research on values have been expanding recently. Cultural exchanges flourish as international relations dictate people's economic and social well-being. Due in part to the importance of the emerging global economy, many countries are rapidly increasing their importance in the international marketplace, and residents of each of these countries may have a slightly different pattern of values, which international marketers will need to fathom.

Cross-national study of values can provide needed cross-cultural understanding of consumers. Japanese marketers, for example, have long promoted their products in the U.S., yet they still have much to learn about the American consumer. One reason Toyota agreed to a joint venture with General Motors was that it provided "a relatively cheap and quick way to learn how to operate in the U.S. with a partner who knows the ins and outs of American business" (Business Week, May 28, 1984, p. 58). American businesses are also eager for information about foreign consumers. In the 1986 NCEIS Trade Analyst readership poll of the top ten trade issues, international marketing was listed as the number one priority. An understanding of cross-cultural values can assist U.S. firms in marketing their products successfully to foreign consumers. The positive impact on profits of individual firms could be accompanied by an improvement in the U.S. foreign trade balance of payments.

Researchers can examine the relation of values to consumption of specific products within different countries. For example, researchers can explore the effect of values on segmentation and targeting of products. Because values influence the way in which consumers react to product offerings, advertising, packaging, pricing, personal selling, and retailing, the effective marketer should be aware of this influence and incorporate it when developing marketing strategy and when planning products. Some failures by firms attempting to penetrate foreign consumer markets may be traceable in part to misunderstanding of values and how these values influence consumer choice

VALS Research

Marketing scholars (e.g., Holman 1984) have informally expressed a great deal of interest in the Values and Lifestyles (VALS) methodology developed at SRI International by Mitchell (1983) and others. The creators of this approach emphasize its combined - segmentation power based on demographics, life style variables, and values. Its conceptualization implies that it has potential for international value research. Researchers have been limited, however, by lack of access to the VALS weighting algorithm, which is proprietary. One purpose of this article is to propose and examine a public algorithm for weighting responses to critical questions that will approximate the VALS weighting system, enabling researchers of international values to use it. This proposed system is based on mirroring empirical data reported by Mitchell (1983).

The development of VALS started from the theoretical base of Maslow's (1954) need hierarchy and the concept of social character (Relsman, Glazer, and Denny 1950). Mitchell (1983) conducted the primary study of VALS. Through statistical and theoretical means he identified attitude and demographic questions deemed useful in classifying people into one of nine life style types. The life style types in the United States include survivors (4%), sustainers (7%), belongers (35%), emulators (9%), achievers (22%), I-am-me (5%), experiential (7%), societally conscious (9%), and integrated (2%).

Weighting of questions for classification was developed using data from a national probability sample of 1635 Americans and their spouses/mates (1078) who responded to an SRI International mail survey in 1980. Although many studies have apparently applied the VALS methodology (Holman 1984), only the 1980 study results have been made public for quantitative inspection, and even then the weighting system was not reported; however, Mitchell did report what questions are used in the VALS algorithm and presented data on the percentage of people who agreed with each item in each VALS type (with the exception of the integrated group3. The lack of public access to the scoring algorithm has also led to a lack of information on the validity of the VALS approach on U.S. data.

We know even less about the degree to which this approach would yield meaningful results in other countries. Mitchell (1983) devotes one chapter to this issue, in which he presents the findings from one VALS study that examines 5 countries. He compares European life styles to American life styles, although he suggests that the research reported is preliminary. Beatty, Kahle, Homer and Giltvedt (1986) utilized the VALS algorithm on a Norwegian sample and found a number of similarities between VALS types in Norway and the United States.

The debate over measuring cross-national consumer values could progress to a higher level if scholars had easy access to a measurement system that approximates VALS. We are proposing that such an algorithm for scoring VALS can be derived. Therefore, the guiding hypothesis for the present study is that it is possible to develop a weighting system that approximates the VALS weighting system.

METHOD

Subjects

The subjects were 167 students enrolled at a large state university in the Western United States. In order to optimize heterogeneity of variance within such a homogeneous group and because of the importance of international values, we drew the primary sample from foreign students who had at least 25 other citizens of their country also enrolled at the University. This limit was applied to ensure that the student at least had an opportunity for ongoing interaction with fellow representatives of his or her culture. One university admission requirement is a score of at least 500 on the Test of English as a Foreign Language (TOEFL), implying respondents have developed at least a minimal facility with English. We also drew a sample of North Americans (70 U. S. citizens and 1 Canadian), again over-sampling out-of-state students. Within these stratification parameters the sampling technique was a simple probability selection. All respondents had lived in the United States for at least 6 months and were thus probably at least minimally familiar with American issues assessed in the survey. Students who failed to reply to the initial mailing received a reminder telephone call the following week and a replacement questionnaire the week after that, resulting in a response rate of 52%.

Materials

Subjects responded to the VALS algorithm items reported by Mitchell. We excluded the questions on political party identification because a large percentage of respondents were not citizens of the United States. We excluded the question on occupation because all respondents were full-time students. We included a direct question on size of home town residence area. Mitchell coded size of residence area from zip codes, but many of our respondents came from countries without zip codes. Finally, we modified the question on household income to personal income because pilot testing revealed that students found the wording of Mitchell's item confusing. That is, they were uncertain whether their household referred to their school or home town household. In several instances where we replicated Mitchell exactly we nevertheless failed to obtain responses in all response categories. For example, in our young sample no one answered the marital status question with widowed.

Categorizing Respondents into VALS Types

Because Mitchell does not supply the weights for computing VALS categories and because it is difficult to gain access to these weights (personal correspondence with B. Warrick, April, 1984), we developed an "empirical mirror" strategy to estimate weights. Mitchell (1983) does provide percentages of his respondents who fell into each VALS group for all but one of the VALS algorithm items. These data served as input for our estimation approach.

To compute the scores for VALS types we first standardized all algorithm items in our data (i.e., the first 25 items in Table 1) except the nominal, bracketed variables of marital status, political outlook, social class, and ethnicity. In accord with the VALS procedure, we calculated 8 separate VALS equations corresponding to each of the VALS types. The national sample percentages reported by Mitchell (1983, p. 279 ff.) were used to compute weighted means and weighted standard deviations for each of the VALS algorithm items. Next, for each item, the difference between the mean for each VALS group and the weighted overall mean was divided by the standard deviation. These numbers served as our weights and are reflections of the degree to which each category varied from the norm on each question in the national sample. The standardized scores of our sample data were then multiplied by the corresponding ratio. For example, the standardized score for item 1 was multiplied by each of the eight group ratios of the form:

group i mean for item 1 - overall mean for item 1

overall standard deviation for item 1    (1)

where group i (1,2,...,8) is each VALS type. This process was repeated for each VALS item except the nominal, bracketed items. At this point, all items referring to a specific group were summed, and this summation was performed for all 8 groups. This can be represented more generally as:

EQUATION

where:

VALS_k = score for empirical mirror VALS group X

S_j = subject's standardized score for item j

X_jk = group k's mean for item j

X_j = overall mean for item j

SD_j = standard deviation for item j.

Next was the treatment of the four nominal, bracketed variables. Because the absolute value of nominal variables has no meaning other than classification into categories, an alternative measure is appropriate to insure that a score is not biased in the wrong direction. Consistent with the previous items, we calculated ratios of the difference between national sample group means and overall means and the national sample standard deviations. This calculation was done for each category level for the four nominal items. For example, for social class, ratios were calculated for each life style and relevant sample category level (3 levels) combination (i.e., 8 x 3 = 24 total ratios). These nominal item scores were then added to the eight life style equations for the level of the nominal variable selected by the respondent, thus yielding eight distinct scores for each subject. Table 1 shows the rounded item weight or bracket weight used for each VALS group for each item or bracket. The eight distinct scores are merely the summation of these rows with each item weight first multiplied by the respective standardized score of the VALS algorithm items, except that the bracketed item weights are added unadjusted. For each subject the eight scores were assessed, and the largest score defined how the subject was classified on VALS. Because Mitchell provided no data for the integrated category, it was not possible to include that category. Also, the item of father's education (but not own education) was excluded because Mitchell provided no data on it. The actual weights in the VALS algorithm are computed at the item-response level.

Because of the common observation among practitioners that VALS relies heavily on demographics, we also computed a demographically adjusted empirical mirror, in which age, income, education, and marital status weights were multiplied by 5.

TABLE 1

WEIGHTS ASSIGNED TO ITEMS FOR VALS CLASSIFICATION

RESULTS

The percentage of the people in each VALS type from Mitchell's national sample and the percentage of people from our sample falling in each VALS category with the actual VALS algorithm, the empirical mirror VALS algorithm, and the demographically adjusted empirical mirror VALS algorithm are presented in Table 2. Only 2 categories differed from Mitchell's national sample by more than 10% for the actual VALS and demographically adjusted empirical mirror: belongers and achievers. The shortage of belongers and achievers probably results from the youth of this sample.

Both of the empirical mirror formulations show a significant relationship with actual VALS. For the unadjusted empirical mirror the relation with actual VALS shows a contingency coefficient of .682, 4 = .932, and a x2(49) = 145.2, p < .001. For the demographically adjusted empirical mirror relation with actual VALS the contingency coefficient is .755, t= 1.15 and x2(49) = 220.87, p < .001. These results imply a strong relation between the empirical mirror techniques and actual VALS. As expected, the demographically adjusted empirical mirror appears to provide a closer approximation to actual VALS than does the unadjusted empirical mirror.

TABLE 2

PERCENTAGE OF RESPONDENTS IN EACH VALS TYPE WITH THE THREE MEASURING SYSTEMS

DISCUSSION

Methodology and International Value Research

Values are difficult to measure even within homogeneous cultures. Much recent research on consumer values has emphasized methodological concerns (e.g., Beatty, Kahle, Homer and Misra 1985; Homer and Kahle in press; Kahle, Beatty, and Homer 1986). Values are among the most abstract of social cognitions (Kahle 1984), thus rendering them elusive to highly concrete measure. Most values are viewed positively by most people, thus leading to positively bias and ceiling effects. Irrelevant needs, salience, super-ordination, impression management, social change, excessive abstractness, and ambiguity of meaning may all distort self-reports of values.

A number of additional methodological challenges exist with cross-cultural surveys and research. Researchers must convey the meaning of questions accurately and in a way respondents can grasp. They must correctly incorporate subtleties and nuances of language. Comparative questionnaires should be translated and back translated by bilingual nationals. Researchers also must select the most effective method of communication (e.g., mail, telephone, or personal interview) within a culture. Construction of an adequate sample is potentially far more complex for researchers in some countries than in others, because sources similar to the ones used to describe populations in some countries may not be available in others. Even with adequate questionnaires, people in some cultures may simply think in ways too different to isolate within a conventional research context. Finally, certain questions to which some cultures are willing to respond may be considered sensitive or inappropriate by others. The generality of the entire approach of cross-national research still needs further demonstration in a variety of nations (the emic-etic problem applied to methodology).

Problems With VALS

The VALS methodology may be particularly prone to some of the above problems, as well as having its own unique problems. Both the collection and analysis of data are a fairly tedious process. The translation and back-translation of 34 items (which is the short form of VALS) along with the various other questions asked in the questionnaire will be time consuming and can lead to a good deal of cultural bias. The meanings and nuances of a set of attitudinal and behavioral statements will simply have a greater chance of distortion and ambiguity than simpler statements. Some of the statements seem particularly tied to the U.S. culture and need considerable modification. Earlier we noted some necessary changes we had to make in the instrument to make it fit our sample. This type of fitting might be necessary for each sample taken. For example, the new algorithm item, "Just as the Bible says, the world literally was created in six days," may simply confuse a Buddhist in Japan.

We are also concerned that the VALS algorithm is heavily dependent upon demographics. It was necessary for us to adjust our demographic questions, and this fact suggests that the VALS questionnaire would always require some adjustment if the sample is demographically non-representative in some way. Further, the meaning of demographics may vary from country to country. For example, the variance in income is much smaller in Norway than in the U.S., suggesting income contributes less to individual differences among Norwegians. VALS' heavy dependence on demographics suggests it is not a pure measure of values or psych( graphics, which reduces the theoretical importance of the methodology and conceptualization.

An Alternative to VALS

One alternative to VALS is the List of Values (LOV), which was developed by researchers at the University of Michigan Survey Research Center (Kahle 1983; Veroff, Douvan, and Kulka 1981). LOV was developed from a theoretical base of Feather's (1975), Maslow's (1954), and Rokeach's (1973) work on values, in order to assess adaptation to various roles through value fulfillment. It is tied most closely to social adaptation theory (Kahle 1983, 1984). Subjects see a list of nine values, including self-respect, security, warm relationships with others, sense of accomplishment, self-fulfillment, sense of belonging, being well respected, fun and enjoyment in life, and excitement. These values can be used to classify people on Maslow's (1954) hierarchy, and they relate more closely to the values of life's major roles (i.e., marriage, parenting, work, leisure, daily consumption) than do the values in the Rokeach (1973) Value-Survey (Beatty et al. 1985) or in VALS (Kahle, Beatty and Homer 1986). In the LOV method, subjects have been asked to identify their two most important values (Kahle 1983; Veroff et al. 1981), to rank the values (Beatty et al. 1985, Kahle et al. 1986), or to rate the values on a 1 to 9 interval scale from extremely important to extremely unimportant (Kahle and Homer in press).

Although LOV and VALS share some disadvantages, the advantage of LOV over VALS are numerous (Kahle et al. 1986). LOV significantly predicts consumer behavior trends more often than does VALS. LOV is easier and quicker to administer. It involves minimal computer work to analyze the findings. It is not as subject to individual interpretation, which can be a considerable problem in international research. That is, the exact phrase may be more easily retained or translated. This advantage also implies that advertising implications may be more obviously based on the findings. LOV can employ higher level (interval) statistics, also.

CONCLUSIONS

These results seem not to disprove the hypothesis, which postulates that it is possible to develop a weighting scheme to approximate VALS. The results suggest a degree of face validity for the "empirical mirror" technique of deriving weights for VALS scoring, especially when demographically corrected. The phi coefficients and other measures of association suggest empirical correspondence. Of course, the empirical mirror weights are not identical to the weights used by SRI International. But small variations in weights may not alter composite construction to any large degree (Wainer 1976; Wang and Stanley 1970). A further limitation is that SRI International periodically revises the actual weights, thus making the exact algorithm somewhat different from time to time.

Additionally, this paper addressed the many methodological problems associated with the study of values and with the use of VALS specifically. An alternative approach, LOV, is presented and compared with VALS on a number of dimensions, such as ease of use and analysis, predictive ability and interpretive/application ability. LOV is found to be superior to VALS on these dimensions.

One of the most fundamental principles of science is public sharing of information (Ziman 1968). "Objectivity and logical rationality, the supreme characteristics of the Scientific Attitude, are meaningless for the isolated individual; they imply a strong social context, and the sharing of experience and opinion" (Ziman 1968, p. 144). The time has come to allow scientists to share research on VALS, especially because many marketing texts now promote VALS. Scientific research on VALS should be encouraged. We are just beginning to unlock important knowledge on consumer values. Which of several methodologies is most useful in marketing research can best be answered by further empirical investigation (Kahle 19853.

REFERENCES

Beatty, Sharon E., Lynn R. Kahle, Pamela Homer and Anders Giltvedt (in press), "A Cross-Cultural Exploration of the VALS Typology," in Robert E. Pitts, Jagdish Seth, and Umberto Valencia, Eds., Cultural and Subcultural Influences in Consumer   Behavior, forthcoming.

Beatty, Sharon E., Lynn R. Kahle, Pamela Homer, and Shekhar Misra (1985), "Alternative Measurement Approaches to Consumer Values: The List of Values and the Rokeach Value Survey," Psychology and Marketing, 2 (3), 181-200.

Becker, Boris W., & Conner, P. E. (1981), "Personal Values and the Heavy Users of Mass Media," Journal of Advertising Research, 21, 37-43.

Berrien, F. K. (1966), "Japanese and American Values," International Journal of Psychology, 71, 129-142.

Feather, Norman T. (1975), Values in Education and Society, New York: Free Press.

Holman, Rebecca (1984), "A Values and Life Styles Perspective on Human Behavior," in Personal Values and Consumer Psychology, Robert E. Pitts, Jr., & Arch G. Woodside, eds. Lexington, MA: Lexington Books.

Homer, Pamela M., & Kahle, Lynn R. (in press), "A Structural Equation Test of the Value-Attitude Behavior Hierarchy," Journal of Personality and Social Psychology, forthcoming.

Kahle, Lynn R., Ed. (1983), Social Values and Social Change: Adaptation to Life in America, New York: Praeger.

Kahle, Lynn R. (1984), Attitudes and Social Adaptation: A Person-Situation Interaction Approach, Oxford, U.K.: Pergamon.

Kahle, Lynn R. (1985), "Social Values in the Eighties: A Special Issue," Psychology and Marketing, 2 (Winter), 231-237.

Kahle, Lynn R. (1986), "The Nine Nations of North America and the Value Basis of Geographic Segmentation," Journal of Marketing, 50 (April), 37-47.

Kahle, Lynn R., Beatty, Sharon E., & Homer, Pamela M. (1986), "Alternative Measurement Approaches to Consumer Values: The List of Values (LOV) and Values and Lifestyle Segmentation (VALS)," Journal of Consumer Research, 13 (December), 405-409.

Kahle, Lynn R., & Eisert, Debra C. (1986), "Social Values in the American Workplace." In Eric Flamholtz, Yvonne Randle, and Sonja Sackmann, Eds., Future Directions in Human Resource Management, Los Angeles: UCLA Publications, 203-223.

Maslow, Abraham H. (1954), Motivation and Personality, New York: Harper.

McQuarrie, Edward F. and Langmeyer, Daniel (1985), "Using Values to Measure Attitudes Toward Discontinuous Innovations," Psychology and Marketing, 2 (Winter), 239-252.

Mitchell, Arnold (1983), The Nine American Life Styles, New York: Warner.

Pitts, Robert E., Jr., and Arch G. Woodside, Eds. (1984), -Personal Values and Consumer Psychology. Lexington, MA: Lexington Books.

Riesman, David, Nathan Glazer, and Revil Denney (1950), The Lonely Crowd, New Haven, Conn.: Yale University Press.

Rokeach, Milton (1973), The Nature of Human Values, New York: Free Press.

Veroff, Joseph, Elizabeth Douvan, and Richard A. Kulka (1981), The Inner American, New York: Basic Books.

Wainer, Howard (19-76), "Estimating Coefficients in Linear Models: It Don't Make No Nevermind," - Psychological Bulletin, 83 (March), 213-217.

Wang, M. W. and Julian C. Stanley (1970), "Differential Weighing: A Review of Methods and Empirical Studies," Review of Educational Research, 40, 663-705.

Zavalloni, M. (1980), "Values," In Harry C. Triandis and Richard W. Brislin, Eds. Handbook of Cross-Cultural Psychology, Vol.5. Social Psychology. Boston: Allyn and Bacon.

Ziman, John (1968), Public Knowledge: An Essay Concerning the Social Dimension of Science, Cambridge, UK: Cambridge University Press.

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