Exploring the Nomological Validity of Life Style Types

John L. Lastovicka, University of Kansas
E. H. Bonfield, Temple University
ABSTRACT - The nomological validity of a set of life style types was tested. Theoretically predictable relationships between life style type and choice of marital partner and between life style type and sex role were empirically evaluated. Modest support for the nomological validity of the life style types was found.
[ to cite ]:
John L. Lastovicka and E. H. Bonfield (1980) ,"Exploring the Nomological Validity of Life Style Types", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 466-472.

Advances in Consumer Research Volume 7, 1980     Pages 466-472

EXPLORING THE NOMOLOGICAL VALIDITY OF LIFE STYLE TYPES

John L. Lastovicka, University of Kansas

E. H. Bonfield, Temple University

[The research reported in this paper was supported by a faculty grant-in-aid of research awarded to E. H. Bonfield by Temple University.]

ABSTRACT -

The nomological validity of a set of life style types was tested. Theoretically predictable relationships between life style type and choice of marital partner and between life style type and sex role were empirically evaluated. Modest support for the nomological validity of the life style types was found.

INTRODUCTION

Consumer researchers have an interest in the reliability and validity of life style measures. For example, one of the earliest published life style studies (Wilson 1966) included reliability estimates for a series of life style trait scales. More recently, researchers appear to be especially concerned with the construct validity of life style types or psychographically defined market segments (Wells 1975, Mehrotra and Wells 1977). As articulated by Wells:

From a policy point of view, the most immediately pressing question concerns the reality of psychographic segments. Marketers who make important decisions on the basis of segmentation studies urgently need ways to determine when the products of cluster analysis or Q factor analysis represent real groups of real consumers, and when they represent figments of the computer's imagination (1975, p. 207).

This concern as to whether consumer life style segments represent meaningful reality as opposed to methodological artifact may be appreciated by reviewing the inductive analytical steps typically used in the construction of life style types. Green and Carmone (1978) have outlined such a sequence of steps which assumes responses have been collected from a sample of n individuals on m life style items.

1.  Correlations are computed for all pairs of m life style items and are factor analyzed.

2.  The retained factors are rotated to facilitate their interpretation.

3.  Based on the content of the life style items with high loadings on each factor, factors are interpreted and given some trait name such a "Swinger Orientation" or "Traditionalist."

4.  Factor scores or sum score measures of each trait are computed for each individual.

5.  A dissimilarity measure, such as a distance measure, is computed for each pair of respondents using the factor or sum score measures of the life style traits. Individuals in the sample are then clustered.

6.  After deciding what number of clusters to use, each cluster is described in terms of its set of average scores on the trait measures. Interpretations given each cluster are meant to correspond to psychographically defined market segments or life style types.

7.  The life style types are then cross tabulated against other descriptions of the individual respondents, such as demographic measures, to see how the life style types differ on additional variables not used in the clustering.

Although the life style types which emerge from such a complex analysis sequence are usually plausible, it typically takes more than face validity to ensure a set of scientifically valid concepts.

BEYOND FACE VALIDITY

Campbell (1960) has specified that construct validity, or the degree to which a concept is what it purports to represent, is composed of both trait validity and nomological validity. Trait validity, and its better known components of convergent and discriminant validity, primarily examine the adequacy of the measurement of a concept, but not the adequacy, or the theoretical meaning, of the concept itself. For example, discriminant validity of a given life style cluster would be supported if the type had a pattern of average scores on factor measures unique from average scores on other types. A clearly unique centroid would discriminate the type in question from other life style types. Nomological validity, however, examines the adequacy of the concept itself as determined by confirmation of theoretically predicted relationships between the measure of the concept in question and the measures of other concepts. Since life style types are the end result of an inductive-empirical sequence, the adequacy of the types as meaningful behavioral concepts--which are more than data s, summaries--needs to be investigated.

Although researchers should be aware of the need for construct validity in life style typologies, little work appears to have been done. One exception is the Darden and Perreault (1976) study in which both life style trait measure reliability and clustering stability were examined. The trait measures were evaluated using split-half reliability measures while cluster stability was examined through replication of the clustering solution on subsamples of the data. It should be recognized, however, that these reliability and stability analyses represent no more than tests for internal consistency. That is, reliability indicates that the items used to compute a sum score scale were multiple measures of the same trait. Further, the cluster cross-validation was simply an attempt to test the hypothetical typology suggested by their empirical analysis in an "independent'' set of data. Clearly, developing and testing a hypothesis in the same set of data is scientifically ,m-tenable. Although using a subset of the hypothesis generating data for confirming life style types is open to question, it nonetheless provided some evidence that consistently distinct life style types were in the data.

Testing for internal consistency provides direct evidence about whether (1) there are underlying life style types which are responsible for the covariation in a set of items and (2) true, systematic, across-person variation (as opposed to random variation) is responsible for the empirically derived typology. However, internal consistency analysis does not help in the specification of meanings of each type, or what each type represents. That is, following Campbell (1960), finding internal consistency provides a partial indicator of trait, but not nomological validity. Accepting the assumption that measures of internal consistency are indicators of reliability, it can also be seen that the reliability of a measure places an upper bound on the degree of relationship between the measure and some other variable (Bohrnstedt 1970). Thus, prior to assessing the convergent and discriminant validity of a typology, high internal consistency ought to be present.

While most researchers appear to have at least informally examined the internal consistency of their life style typologies, none apparently have tested for convergent or discriminant validity. Since consumer researchers typically measure life style utilizing only respondents' subjective self reports of activities, interests, and opinions, there has been little opportunity to test for convergent validity of life style constructs. For example, if two maximally different methods of life style type classification--e.g., subjective self report and a reputational evaluation by a judge or expert--were used, convergent validity could be assessed. Except for the recent Szybillo, Binstok, and Buchanon (1979) study in which convergent validity for a few activity items was examined, life style researchers appear to assume respondents are totally truthful in representing themselves on AIO inventories. As shown in personality trait assessment, respondent faking is a potentially severe problem for research based excessively on self reports (Anastasi 1968).

There is evidence that discriminant validity of life style types has been at least intuitively assessed by some consumer researchers. Claims that life style market segments differ from demographically defined market segments or analyses which detect uniqueness of life style cluster centroids hint of discriminant validity. Discriminant validity is supported whenever there is evidence that a life style typology does not correspond with concepts with which it should not correspond.

Very clearly, however, there has been little, if any, assessment of the nomological validity of life style types. Although the analytical steps which lead to a life style typology have the virtue of an almost unquestioned tradition, the adequacy of the deductions which can be drawn from the description of life style types is unknown. That is, a meaningful life style typology should be usable as a theoretical concept in a network of theoretical concepts to help generate testable propositions. While relationships involving types have been found (e.g., Wells has reported "he-man" types to be heavy beer drinkers), end while such findings are apparently replicable (e.g., Kinnear and Taylor 1976), they appear to have been originally post hoc discoveries in life style data bases.

PURPOSE OF THE STUDY

The purpose of the current research has been to explicitly test the nomological validity of a set of empirically defined male and female life style types. Propositions about relationships between the life style types and a pair of concepts infrequently found in consumer research, i.e., marital partner choice and sex role orientation, have been examined. Nomological validity for the life style types is enhanced to the degree that the following two propositions are supported.

Proposition 1: Likes Attract

The likelihood of marriage between pairs of empirically determined male and female life style types will be inversely proportionate to the degree of difference of the types in their activities, interests, and opinions. This proposition examines the notion that similar life style types will tend to marry, i.e., that opposites do not attract. Specifically, using the life style types reported by Mehrotra and Wells, the proposition predicts that "Herman, the retiring homebody" would be more likely to wed "Thelma, the old fashioned traditionalist" than "Candice the chic suburbanite."

Proposition 2: Type and Sex Role

Both male and female life style types should be distinctive in terms of sex role orientation. Borrowing, again from the Mehrotra and Wells typology, this proposition predicts Thelma would be as feminine as Candice, but Candice would be more masculine than Thelma. Therefore Candice would be either androgynous or masculine.

Theory is only sketchy with respect to both of these propositions. In support of the "likes attract" hypothesis, Berelson and Steiner (1964) have reported that not only do people tend to marry within their own and adjacent social classes, but also, following Roth and Peck (1951) adjust better to married life. In addition, Kerchoff (1977) has shown an overwhelming majority of U.S. citizens marry within their own ethnic (racial, religious, national origin) groups. Since Myers and Gutman (1974) have shown small, but consistent and significant, relationships between social class and life style measures, it seemed reasonable to extend the "likes attract" notion to life style types.

The second proposition is supported by the observation that life style researchers seem to seek personality traits or typologies more directly related to standard, more global personality traits or typologies (Kassarjian 1971, Wells 1975). Following Allport's (1961) hierarchical notion of gradations, or levels, of traits, life style traits have implicitly been viewed by consumer researchers as secondary dispositions operant only in limited behavioral situations such as the acquisition and/or consumption of consumer goods. These secondary, or life style, traits are influenced by more general traits called cardinal or central traits such as masculinity and femininity. Thus, just as Greeno, Whipple, and Kernan (1973) were able to uncover and provide post hoc explanations for the relationships between a set of life style types and four Gordon Personality Profile traits, it should be possible to test a priori relations between sex role orientation and a life style typology.

The theoretical development supporting the two propositions is admittedly weak. However, as Chronbach and Meehl (1955) have pointed out, nomological validation in the behavioral sciences typically, end of necessity with new concepts, works from theories that are sketches as opposed to well-formed, interconnected sets of theoretical concepts. The purpose of early nomological validation studies such as this one is to provide tests of these theoretical concepts in order to develop more well formed, interconnected theories. As the number of studies testing a particular concept increases, the level of conceptual rigor and sophistication of these studies is also expected to increase leading to better specified theoretical networks and more rigorously evaluated concepts.

METHODOLOGY

Separately self-administered questionnaires were successfully completed by both spouses in 98 households selected by means of a cluster sample (Sudman 1976). Clusters of blocks were drawn from census tracts of a middle to upper class suburb of a large northeastern metropolitan city. The bulk of the questionnaire completed by the 196 respondents contained questions regarding family decision making across a large set of household goods and services. The remainder of the questionnaire consisted of batteries of sex role, life style, and demographic questions.

TABLE 1

THE MASCULINE AND FEMININE BSRI ITEMS

The complete Bem Sex Role Inventory (1974, identified as BSRI) was used to assess sex role for this study. The BSRI contains both a masculinity and a femininity scale, each with 20 personality characteristic items selected on the basis of sex-typed social desirability. The set of 40 masculine and feminine items is shown in Table 1. Respondents used a 7-place scale with 1 and 7 anchored at "never or almost never true" and "always or almost always true" respectively to indicate the degree to which the items were appropriate as a self rating. The BSRI was developed to preclude "an inverse relationship between masculinity and femininity (Bem 1974, p. 155," and tests have shown the intended logical independence to be demonstrable. Bem has shown the scores of the masculinity and femininity scales to be basically uncorrelated (r = .05). On the basis of respondents' scores on the two scales, each respondent receives an androgyny score which is defined as the difference between the femininity and masculinity scores expressed as Student's t statistic. That is, the androgyny scale reflects the difference between means on the masculine and feminine items divided by a term reflecting the variances of the two sets of ratings of the individual respondent. Lower scores on the androgyny scale are indicative of a higher state of androgyny while higher scores on the feminine and masculine scales are indicative of higher states of masculinity and femininity respectively. Cronbach alpha reliability assessments of the three scales have been consistently high with values ranging from .80 to .86.

The life style questions consisted of a battery of 21 different life style items. Each item was marked on a 1 to 6, "definitely disagree" to "definitely agree" scale. Following the analytical sequence suggested by Green and Carmone (1978), five sum-score life style scales were empirically developed. These life style scales are shown in Table 2 along with the items constituting each scale. These scales were developed pooling all 196 male and female respondents.

TABLE 2

LIFE STYLE SCALES

Veldman's (1967) hierarchical clustering procedure, HGROUP, was used on the 91 men without missing scale data to form five male life style types, while separate analysis of the 91 women without missing data--again using HGROUP--lead to seven female life style clusters. For both male and female clustering, the data consisted of measures on five standardized sum score scales. In line with the standard procedure, these clusters were interpreted on the basis of their centroids in the five dimensional life style space. Interpretations of these life style types are shown in Tables 3 and 4.

TABLE 3

MALE LIFE STYLE SEGMENTS

TABLE 4

FEMALE LIFE STYLE SEGMENTS

Five psychometric specifics of this empirically derived life style segmentation should be listed. First, the five factor principal axes factor solution explained 74% of the variance. Second, the sum-score scales were found to be reasonably orthogonal, the highest correlation being -.17 between the "Positivism-Self Confidence" and "Frustration" scales. Third, the reliability statistics for each scale shown in Table 2, although not as high as standardized scales such as the BSRI, have a range (.50 to .80) presented as acceptable in most published life style research. Fourth, the clustering solution produced groups that were unique in that analyses of variance end post hoc multiple comparison tests showed each life style type to be significantly differentiated from other life style types on one or more of the five sum-score scales. Finally, the last drastic change in the marginal sum of squared distances occurred with the males when going from the four to five group solution and with the females when going from the six to seven group solution.

TESTING THE NOMOLOGICAL PROPOSITIONS

The Likes-Attract Proposition

The first proposition argues that if life style types are real, like life style types ought to disproportionately attract each other and marry. Thus, between pairs of female and male types, there should be an inverse relationship between difference in life style and likelihood of marriage. Differences in life style between females of Type i and males of Type j were operationalized by computing the Euclidean distance between the centroids of types i and j in the five-dimensional life style trait space. Likelihood of marriage was operationalized as the observed proportion of type i women married to type j men.

FIGURE 1

PLOT OF PROPORTION OF FEMALE TYPE I MARRIED TO MALE TYPE J BY EUCLIDEAN LIFE STYLE DISTANCE BETWEEN FEMALE TYPE I AND MALE TYPE J

The plot of the proportions versus the distances is shown in Figure 1. The significant correlation between the two variables, r = -.49, provides modest support for the first proposition.

The Type and Sex Role Proposition

Although the first proposition could be tested in the abstract) i.e., without descriptions of the life style types, the second proposition demands that specific inferences be drawn for each of the 12 life style types described in Tables 3 and 4. By attending especially to the "Positivism-Self Confidence" and "Home-Family Orientation'' descriptions in each life style type, a relative sex role orientation was predicted, a priori, for all 12 life style types. These predicted relative sex role orientations shown in Table 5, describe those life style types that ought to be more masculine, feminine, or androgynous than the remaining life style types of the same sex. These expectations were used as a basis for computing multiple comparison tests between the life style types on the masculine, feminine, and androgyny scales. Thus, for example, the sex role expectations regarding masculinity for the male types correspond to the statistical test of the hypothesis that male types 1, 3, and 4 ought to score higher on the BSRI masculinity scale than male types 2 and 5. At the same time) male type 2 should score more androgynous than any of the other four male types.

Table 6 contains the means of the different life style types on the BSRI scales as well as the results of statistical tests for type differences. As the ANOVA F statistics demonstrate, significant differences were found between the male types on the masculine scale and between the female types on the masculine and androgyny scales. Where there were both sex role predictions and significant overall differences between the types, multiple comparison tests were executed. As expected, male types 1, 3, and 4 were more masculine than male types 2 and 5; and female types 2, 6, and 7 were more masculine than female types l, 3, 4, and 5. These results suggest support for the second proposition. That is, while the predictions about femininity and androgyny were not supported, the predictions regarding masculinity appear to be sound.

TABLE 5

RELATIVE SEX ROLE ORIENTATION PREDICTIONS ACROSS THE LIFE STYLE TYPES

TABLE 6

MEANS ON THE BSRI MASCULINE, FEMININE, AND ANDROGYNY SCALES BY LIFE STYLE TYPES

DISCUSSION

The pursuit of nomological validity in the case of the likes-attract proposition is diagramed in Figure 2.

FIGURE 2

A SCHEMATIC REPRESENTATION OF THE RATIONALE OF NOMOLOGICAL VALIDITY

The figure shows that the link of interest, link 1, can only be assessed by proxy, that is, by assessing link 4. Using link 4 as evidence for link 1 can only be done by assuming adequate trait validity in links 3 and 2. In short, the information gained about link & is only as good as the assumptions about links 3 and 2. As suggested by link 5, modest link 1 relationships may result from artificial reductions due to other factors which have not been dealt with by the research design or analysis. Thus, link 5 in Figure 2 suggests mutual physical attraction may influence the likelihood of marriage in addition to similarity of life style type. It is possible that if the effect of link 5 could be removed, the degree of relationship in link 1 would increase. The dilemma suggested by Figure 2 is that an investigator interested in evaluating the validity of life style types cannot investigate everything at once. Specifically, nomological validity can only be investigated by assuming trait validity as well as assuming all important concepts have been taken into account.

Consider the psychometric properties of the operational definitions used in the current study. Bem's (1974) BSRI measures are standardized and have known properties. The ad hoc life style measures, relative to those reported in the literature, appear adequate. Thus, the assumption of trait validity leads to only modest support for nomological validity for the specific life style types found in this research. Nonsupport for nomological validity is particularly troublesome when it is considered that the two propositions tested represent common sense notions. If the life style concept is too weak to map nearly obvious relations in a theoretical network, its usefulness may be overrated. This observation should not be construed as an attack on the life style concept or previous descriptions and discussions of life style types. Indeed, these results point out the need for more time to be spent on nomological validation of the life style typology developed in this study. As demonstrated here, even when life style measures have apparent trait validity, reasonable and logical inferences may only be modestly supported. Such circumstances will continue to raise the question: "Are these life style types real?"

One of the outcomes of early nomological validation studies should be enrichment of the theoretical network in which the concepts to he validated are nested. It is hoped the following notions will provide for relationships useful in addressing the reality of life style types question.

First, Allport's notion of a hierarchy of personality dispositions appears modestly supported. In the figure below, link 1 was examined and partially supported. At the same time, previous research has established relations shown by link 2. If there is now

DIAGRAM

increased knowledge of the sex role characteristics of the life style types, there should also be increased ability to predict consumption behavior characteristics of the different life style types. For example, further inspection of Table 6 suggests masculinity is the key differentiating trait between life style types of both sexes. That is, the data suggests that although both males and females are distinguished by equal and fixed levels of femininity, there is no within-sex differentiation in femininity. Across female types, there is more of a difference in masculinity than femininity. Thus, it is reasonable to expect that purchasing decision differences between the female life style types would be manifested more in terms of traditionally male products like lawnmowers and automobiles than traditionally female products like household cleaning products or cosmetics. Note that this developing proposition involves both inbound links (1, above) to the concept in question and outbound links (2, above). This represents a slightly more complex proposition then either Proposition 1 or 2 which dealt with outbound or inbound links only.

Second, regarding the likes-attract proposition, it appears that "birds of a feather" do tend to "flock together.'' However, this similarity is hardly a one-to-one correspondence on each trait. Thus, there is more household life style information variation with husband and wife life style measures than with either measure alone. Considering the success in using individual life style measures in explaining what essentially is household behavior (e.g., Plummet 1971), at least additive, if not an interactive, increments should be expected when explaining household behavior with Joint life style information. Thus, steps should be taken to develop a family life style conceptualization.

Of necessity, behavioral science develops by a cyclical process with observation leading to concepts, and concepts guiding the next set of observations. In life style research, a multitude of items have been linked to traits which, in turn, have been used to define life style typologies. The almost lay-like typologies have all too often not been linked in a critical manner to other behavioral concepts. The deductive half of the inductive-deductive research cycle should be addressed in life style research.

REFERENCES

Allport, G. W. (1961), Pattern and Growth in Personality, New York: Holt, Rinehart & Winston.

Anastasi, A. (1968), Psychological Testing, London: MacMillan Co.

Bem, S. L. (1974), "The Measurement of Psychological Androgyny," Journal of Consulting and Clinical Psychology, 42, 155-62.

Berelson, B. and Steiner, G. A. (1964), Human Behavior: An Inventory of Scientific Findings, New York: Harcourt, Brace & World.

Bohrnstedt, G. W. (1970), "Reliability and Validity Assessment," in Attitude Measurement, ed. G. P. Summers New York: Rand McNally.

Campbell, D. T. (1960), "Recommendations for APA Standards Regarding Construct, Trait and Discriminant Validity,'' American Psychologist, 15, 546-53.

Cronbach, L. J. and Meehl, P. E. (1955), "Construct Validity in Psychological Tests," Psychological Bulletin, 52, 281-302.

Darden, W. R. and Perreault, W. D. Jr. (1976), "Identifying Interurban Shoppers: Multiproduct Purchase Pattern and Segmentation Profiles," Journal of Marketing Research, 13, 51-60.

Green, P. E. and Carmone, F. J. (1978), "Some Methodological Alternatives in the Analysis of Life Style Data," Journal of Business and Economics, 30, 158-61.

Greeno, D, W., Sommers, M. S., and Kernan, J. B. (1973), "Personality and Implicit Behavior Patterns," Journal of Marketing Research, 10, 63-9.

Kassarjian, H. H. (1971), "Personality and Consumer Behavior: A Review," Journal of Marketing Research, 8, 39-43.

Kerchoff, A. C. (1977), "Patterns of Marriage and Family Formation and Dissolution," in Selected Aspects of Consumer Behavior, ed. R. Ferber, Washington: National Science Foundation.

Kinnear, T. C. and Taylor, J. R. (1976), "Psychographics: Some Additional Findings," Journal of Marketing Research, 13, 422-5.

Mehrotra, S. and Wells, W. D. (1977), "Psychographics and Buyer Behavior: Theory and Recant Findings," in Consumer and Industrial Buying Behavior, eds. A. G. Woodside, J. N. Sheth, and P. D. Bennett, New York: North-Holland, 49-65.

Myers, J. H. and Gutman, J. (1974), "Life Style: The Essence of Social Class," in Life Style and Psychographics, ed. W. D. Wells, Chicago: American Marketing Association.

Plummet, J. T. (1971), "Life Style Patterns and Commercial Bank Credit Card Usage," Journal of Marketing, 35, 35-41.

Roth, 3. and Peck, R. F. (1951), "Social Class and Social Mobility Factors Related to Marital Adjustment," American Sociological Review, 16, 478-87.

Runkel, P. J. and McGrath, 3. E. (1972), Research on Human Behavior: A Systematic Guide to Method, New York: Holt, Rinehart & Winston.

Sudman, S. (1976), Applied Sampling, New York: Academic Press.

Szybillo, G. J., Binstok, S., and Buchanan, L. (1979), "Measure Validation of Leisure Time Activities: Time Budgets and Psychographics," Journal of Marketing Research, 16, 74-9.

Veldman, D. J. (1967), Fortran Programming for the Behavioral Sciences, New York: Holt, Rinehart & Winston.

Wells, W. D. (1975), "Psychographics: A Critical Review," Journal of Marketing Research, 12, 196-213.

Wilson, C. L. (1966), "Homemaker Living Patterns and Marketplace Behavior: A Psychometric Approach," in New Ideas for Successful Marketing, eds. J. S. Wright and J. L. Goldstucker, Chicago: American Marketing Association.

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