Validation of the Cad Instrument: a Replication

ABSTRACT - A replication study involving the CAD personality instrument was conducted. Adequate reliability and moderate validity estimates were demonstrated. However, results indicated a weaker internal structure of the CAD measure.

Citation:

Pradeep W. Tyagi (1983) ,"Validation of the Cad Instrument: a Replication", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 112-114.

Advances in Consumer Research Volume 10, 1983      Pages 112-114

VALIDATION OF THE CAD INSTRUMENT: A REPLICATION

Pradeep W. Tyagi, San Diego State University

ABSTRACT -

A replication study involving the CAD personality instrument was conducted. Adequate reliability and moderate validity estimates were demonstrated. However, results indicated a weaker internal structure of the CAD measure.

INTRODUCTION

The purpose of this study is to assess the reliability and validity of Cohen's (1967) personality instrument designed to measure consumers' interpersonal orientations--Compliance, Aggression, and Detachment. The study is a replication of Noerager's (1979) study, which was aimed to assess the validity of the CAD instrument. A replication study is defined as a study which duplicates only the sampling and experimental procedures of the first study (Lykken 1978). As such, this research is a deliberate attempt to follow Kollat, Engel, and Blackwell's (1970) and Jacoby's (1978) appeal for replication in marketing research.

In his original study, Cohen (1967) demonstrated the nomological validity of the CAD scale by predicting several aspects of consumer behavior. Since the CAD instrument offers a simple and inexpensive means to obtain data or personality-related measures, its use was recommended for predicting the buying decisions relating to the products having interpersonal significance. Noerager's study (1979) raised serious questions about the reliability and validity of the CAD instrument. The reliability coefficients and validity estimates reported were significantly lower than the arbitrary minimum standards. Factor analytic scores indicated a weak internal structure of the CAD scale.

Though Noerager's study was operationally sound, it had the following shortcomings: (1) the subjects were army trainees and not consumers whom the original scale was intended for; and (2) the 16 PF instrument, used to assess the convergent and discriminant validities, is not designed to measure the interpersonal-orientation personality variables as specified in Cohen's instrument (a limitation clearly identified in Noerager's study, p. 56). Steps are taken in the present study to overcome the above limitations and hence provide a more meaningful assessment of the CAD instrument. Specifically, this replicative research was designed to assess:

(a) the CAD's internal consistency reliability;

(b) the CAD's convergent and discriminant validities; and

(c) the CAD's internal structure.

This was accomplished using a random probability sample of consumers and using a criterion measure conceptually and operationally related to the interpersonal-orientation personality variables employed in Cohen's CAD measure.

METHODS

The Sample

The data were collected from a population in a small city located in the southeastern part of the U.S. A random probability technique was used. The subjects were consumers as opposed to students in Cohen's (1967) study or army trainees in Noerager's (1979) study. Only the heads of households who regularly engage in purchasing decision process were asked to complete the questionnaire. Questionnaires were personally distributed and then collected a day later by interviewers. One hundred and seventy questionnaires were distributed of which 148 usable questionnaires were obtained.

Reliability and Validity

To assess the internal consistency of CAD dimensions Cronbach Alpha Spearman-Brown split half and Guttman Lambda coefficients were calculated. The multitrait-multimethod matrix procedure (Campbell and Fisk 1959) was used to examine both convergent and discriminant validities. This procedure examines the extent to which the construct measure correlates with the same construct using dissimilar measures and discriminates from other constructs using similar and dissimilar measures. the criterion measures selected for this procedure are the scales of Nurturance Aggression and Autonomy dimensions from Stern s (1970) Activities Index. The personality variables were defined as follows:

Nurturance: Supporting others by providing love assistance or protection.

Aggression: Disregard for the feeling of others as manifested in hostility either overt or covert direct or indirect.

Autonomy: Detachment from others independence or self reliance.

These scales were specifically selected because of their close affinity to the interpersonal-orientation variables involved in the CAD instrument. The Nurturance construct is conceptually similar to the Compliance construct of the CAD. The Aggression construct is also conceptually akin to the CAD s Aggression construct. And similarly the Autonomy construct is equivalent to the CAD s Detached construct. A panel of three personality psychologists was selected to assist in the examination of conceptual similarity between CAD and Nurturance-Aggression-Autonomy instruments. A consensus was found among the judges with regard to the similarity between two scales. Data indicate that the Nurturance-Aggression-Autonomy scale is characterized by high scale reliabilities (R - R 20) ranging from .51 to .88 (Stern 1970). Test-retest estimates or reliability have also been shown to be adequately high (Walsh 1973). Finally the internal consistency of each scale has shown to be high indicating that the scales are homogeneous (Stern 1970).

Finally a factor analysis was conducted to examine the factor structure of the CAD instrument. The factor analysis procedure has been regarded as an appropriate method to examine similarities among test items with the main dimensions (Kerlinger 1973). A pattern involving three factors each reflecting a different personality variable was expected.

RESULTS

Reliability Estimates

Results from three reliability tests are presented in Table 1. Guttman Lambda coefficients represented the highest reliability estimates. Cronbach Alpha the most commonly used reliability estimate in marketing research showed adequate reliability for the CAD dimensions. Alpha estimates varied from a low of .62 for the Aggressive dimension to a high of .72 for the Compliant dimension. The Alpha coefficients in the present study are within the generally acceptable limits of .5 to .8 (Nunnally 1967). Though the direction of Alpha coefficients paralleled Noerager s (1979) results our coefficients were greater and thus supported the internal consistence of the CAD instrument.

TABLE 1

RELIABILITY COEFFICIENTS FOR CAD DIMENSIONS

Convergent and Discriminant Validity

A multitrait-multimethod matrix was constructed to examine convergent and discriminant validity coefficients of CAD scale. All of the entries in the validity diagonal are significant. Although all entries in validity diagonal are statistically significant they are not high enough to show a clearcut evidence of convergent validity

One of the conditions for the discriminant validity requires that a personality dimension as measured by one method not correlate with other personality dimensions measured by the same method. Table 2 indicates that all three variables do show significant discriminant validities. The second criterion for discriminant validity cakes each diagonal correlation and compares its value to the correlations obtained between each individual trait in the validity diagonal and the other traits when these are measured by the same method (i.e. in the adjacent hetrotrait-monomethod triangles). Except the detached dimension other dimensions meet this criterion satisfactorily. Finally a third criterion for discriminant validity requires that the pattern of correlations within all hetrotrait triangles should be consistent. If for example the compliant-aggressive correlation is greater than the compliant-detached correlation in the first hetrotrait triangle then the compliant-aggression and compliant-autonomy should follow the similar pattern in the second hetrotrait triangles and so on. Clearly this is not the case in hetrotrait triangles presented in Table 2.

Evidently the convergent and discriminant validity test show mixed results at best. However these results indicate a better performance iv the CAD scale than Noerager's (1979) validity findings.

TABLE 2

MULTITRAIT - MULTIMETHOD MATRIX FOR THREE TRAITS AND TWO METHODS

Internal Structure of CAD

The 35 CAD items of the CAD instrument were subjected to a factor analysis using a varimax rotation. (This procedure was used to maintain consistency with Noerager's (1979) analysis.) The CAD instrument consists of 10 items, each to measure compliant and detached dimensions and 15 items to measure the aggressive dimension. Items corresponding to a personality dimension were expected to load on the same factor to indicate a "crystallized" internal structure of CAD. In an attempt to examine the three CAD dimensions structure, a three-factor solution was forced.

Several observations are appropriate in examining the results of three-factor solution in Table 3 (on the following page). First, the items of the compliant dimensions were reasonably related to each other, though factor loadings were only moderate (i.e., less than + .50). This trend was similar to that of Noerager's (1979, p. 57) factor scores. Second, many of the items of the aggressive and detached dimensions tended to load on the same factor. These results indicate a weaker internal structure of the later two dimensions. Third, some items failed to load significantly on any of the extracted-factors. Item 4.of the compliant dimension, for example, did not show a loading of greater than .3 on any of the extracted factors. Fourth, factor loadings of many items were substantially low (i.e., less than + 0.5, a value generally considered arbitrary minimum in the literature). Finally, there was a tendency for some items measuring the same CAD dimension to load on different factors. These results suggest that many of the items may not be consistently measuring the same dimensions and contribute little to the assessment or the three interpersonal orientations. The pattern or factor loadings, therefore, suggests a weak internal structure of the CAD instrument.

CONCLUSIONS

The results of the assessment of the CAD instrument were mixed. Consequently, Noerager's (1979) conclusions regarding the CAD's lack of validity are partly supported. The scale seems to do moderately well on the reliability criterion and concurrent and convergent validity criteria. The performance on the discriminant validity criterion is mixed. The factor loading pattern of 35 CAD items reflects that the internal structure of the CAD instrument is not crystallized. In the light of the above results, further refinements of CAD items (with an appropriate item analysis) to strengthen the internal structure is recommended.

TABLE 3

THREE FACTOR SOLUTION OF CAD ITEMS

REFERENCES

Campbell, D. T. and Fiske, D.W. (1959), "Convergent and Discriminant Validation by the Multitrait-Multimethod Matrix," Psychological Bulletin 56 (March), 81-105.

Cohen, J. B. (1967), "An Interpersonal Orientation to the Study of Consumer Behavior," Journal of Marketing Research, 4 (August), 270-78.

Kerlinger, F. X. (1973), Foundations of Behavior Research, 2nd ed., New York: Holt, Rinehart, and Winston, Inc.

Lykken, D. T. (1978). "Statistical Significance in Psychological Research," Psychological Bulletin, 70, (September), 151-9.

Noerager, J. P. (1979), "An Assessment of CAD--A Personality Instrument Developed Specifically for Marketing Research," Journal or Marketing Research, 14 (February), 53-9.

Nunnally, J. (1967) Psychometric Methods, New York: McGraw-Hill.

Stern, G. G. (1970), People in Context, New York: John Wiley and Sons, Inc.

Walsh, W. D. (1973), Theories of Person-Environment Interaction: Implications for the College Students Iowa City, Iowa: American College Testing Program.

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