Differences Between Intenders and Nonintenders--A Methodological Question


Raymond H. Suh (1972) ,"Differences Between Intenders and Nonintenders--A Methodological Question", in SV - Proceedings of the Third Annual Conference of the Association for Consumer Research, eds. M. Venkatesan, Chicago, IL : Association for Consumer Research, Pages: 512-521.

Proceedings of the Third Annual Conference of the Association for Consumer Research, 1972      Pages 512-521


Raymond H. Suh, California State University, Long Beach

[This study is a part of the author's doctoral dissertation at the University of Illinois, Urbana-Champaign, and the author wishes to express his appreciation to Professor Robert Ferber for providing data and assistance.]

[Raymond Suh is Assistant Professor of Marketing at California State University, Long Beach.]

Buying plans and attitudes toward durable goods have received primary emphasis in the intentions data collected from consumers. Most individual durable goods purchased involved a substantial expenditure, are infrequently purchased by any single household, and provide considerable latitude in the timing of their acquisition. These characteristics provide analytical support for the belief that many households plan their purchases of durables well in advance and are able to offer accurate information about their decisions in interview.

The basic idea behind a survey of consumer buying intentions is that consumer purchases, of items such as houses, automobiles, and appliances, are subject to fluctuations that are in some degree independent of the effects of socio-economic variables. Fluctuations in these forms of expenditures are believed to be more accurately predicted by both the changes in socio-economic variables and consumer pessimism and optimism than by socio-economic variables alone (Juster, 1966, 1964).

One of the problems for an intentions survey is its inefficiency of the basic predictors of purchase rates. As a method of data collection, responses are generally classified into several categories such as "definitely will buy," "probably," and "no," for the expression of buying intentions. The usefulness of the survey is then evaluated by relating variations in the fraction of one or more groups of intenders (respondents reporting "definitely" or "probably") to variations in the fraction reporting purchases. It is puzzling in most studies that data represent only the intenders; and there is no survey evidence that bears directly on the predictability of the critically important movements in nonintenders' purchase rates.

Namias' (1960, 1959) studies indicated that consumers who do not plan to buy household durable goods are more consistent in carrying out their no-buying-intentions than those who plan to buy. On the other hand, a large number of consumers, although small in population, who do not plan to buy,change their minds. Juster (1966, 1964) noted that the majority of durable goods purchases are made by households that are generally classified as nonintenders by consumer surveys. As a consequence, the accuracy of purchase predictions based on intentions surveys depends largely on whether or not changes in proportion of intenders can successfully predict changes in the purchase rate of nonintenders being strongly correlated over time. Piskie (1963) has provided similar arguments in his study. However, Piskie argued that researchers should not only have purchase data on their subjects but should also know subjects' buying plans and the degree of certainty of those plans.

It also appears that buying intentions data are collected, in general, for the wife as a sole respondent who is assumed to represent the family's buying plans. Rarely did studies look into the differences between the husband and wife when contemplating buying intentions, and their subsequent effects on the accuracy of the prediction of the durable goods purchases. Some skepticism, therefore, does exist concerning the method of data collection for family decision making. For example, as early as 1955, Ferber said:

The method by which those studies have been carried out by soliciting the opinion of one family member regarding the purchase influence possessed by himself and by other members of the family and the scarcity of any rigorous tests of the validity provides sufficient basis for receiving such studies with skepticism (Ferber, 1955a, p. 225).

By and large, the findings derived from a number of works are conflicting and thus not very convincing. Some results, based on interview results from a single respondent in a household, suggested that the amount of discrepancy in a household is small when aggregate responses for husbands and wives are compared. This is not so when the responses from husbands and wives in the same families are compared (Blood & Wolfe, 1960; Granbois & Willett, 1970; Heer, 1963, 1962; Sharp & Mott, 1956; Wallin & Clark, 1958). Other disclosures demonstrated that poor agreement between husbands and wives on the family income, durable goods buying plans, actual purchases of the planned buying, and the frequency of sexual intercourse has shown a poor agreement between the members involved (Davis, 1970; Ferber, 1955a, 1955b; Haberman & Elinson, 1967; Suh, 1972).

It appears that these controversies are due to difficulties in data collection methods, Ferber, in two previous works (1955a, 1955b), argued that the reliability of ratings concerning the relative influence of different family members on purchases obtained by direct questioning of only one of the family members is highly limited. Re also indicated that the securing of reliable data, at least in consumer purchase studies, is much more difficult than previously purported. Suh (1972a, 1972b) further indicated that the salient motives for husbands and wives are different for contemplating buying intentions, and more accurate data could be obtained by interviewing respondents during a period varying Prom three to six months and by securing at least two responses from each family.

Therefore, the objective of this study was to determine the differences between intenders (i.e., respondents whose purchase probabilities on living room furniture were substantially high) and nonintenders in estimating purchase probabilities for living room furniture based on four buying intentions scores, eight attitude scores, and eight demographic variables. The cutoff point of intenders and nonintenders was also to be determined. A discriminant analysis was performed for husbands and wives separately, and responses from each sex are compared to examine possible sex differences.


The subjects were members of the Illinois - Berkeley Panel of Consumer Decision Processes who were recently married couples with the husband being 30 years of age or less. The sample size was 230 couples, 166 living in Peoria and 64 in Decatur, Illinois, both cities being in the 100,000 - 200,000 population range. The sample members were selected by systematic random selection from a list of such couples married between June 1, 1968, and September 30, 1968.

Most questions on durable goods purchases and buying intentions were asked of panel members since Fall, 1968. Interviews were conducted only by personal interviews and at intervals varying from two to six months.

Subjective purchase probabilities were used as a measure of durable goods buying intentions. The use of subjective probability had certain advantages over other scaling techniques such as a dichotomous (yes or no) form of the responses, or gradational adjectives (i.e., "definitely will," "probably will," "maybe," "probably not," and "definitely not") in obtaining information on buying intentions. Advantages included: first, the additional information throws light on the degree of certainty with which particular purchase plans may be carried out; second, introduction of buying plans data in this form may serve to further improve the ability to anticipate actual purchases (Ferber & Piskie, 1965).

The purchase likelihood scales were eleven-point scales ranging from "absolutely certain (100%)" to "no chance whatsoever (0%)". The scale appeared as follows:

100% -- Absolutely certain










0% -- No Chance Whatsoever

Only living room furniture, of the sixteen durable products studied, was used in the analysis because living room furniture produced the maximum possible sample size, 230. There were eight demographic variables, eight attitudes scores and four buying intentions scores (See Table 1). Because this was a by-product of a broader study (Suh, 1972a), the study did not provide a process from which both factor scores for attitudinal classification and buying intentions could have been derived.

Once subjective purchase probabilities and rank order information on the sixteen durable goods were obtained, the joint distribution of the rank order information in which each family member hoped to acquire was examined to determine the extent of agreement for living room furniture between husbands and wives. Because the subjective purchase likelihood information was a measure of buying intentions, this measure was used to classify subjects into intender and nonintender groups.


The first task was to determine the cutoff point of intenders and nonintenders on the probability scales. Using the objective criterion, it appeared that the purchase probability 0.5 (50%) would be a reasonable estimation of the cutoff point. Very little empirical evidence was available on this point; however, several researchers (Ferber & Piskie, 1965; Juster, 1964; Piskie, 1963) supported this argument. The selection of the purchase probability 505 as a cutoff point was reasonable in the sense that if a respondent is forced to answer yes or no in regard to a future purchase, the cutoff point on the probability scale would most likely be at .50. Therefore, when the subjective probability is a measure of buying intentions, all those respondents given a .50 probability or above could be placed in the yes category (i.e., intender group) and vice versa.

Once the cutoff point was selected, three questions were of primary interest in this analysis. First, did the two groups occupy different regions in the one-dimensional discriminant space? Secondly, what were the discriminating variables? Thirdly, how well could we explain the classification of subjects into the intender and nonintender groups?

Discriminant Analysis for husbands: Out of the total 218 subjects (Table 1), 93 husbands were classified as intenders and 125 as nonintenders. The intender group was operationally defined as the group of subjects whose purchase probabilities were equal to or greater than 0.50; and the nonintender group was the group of people whose purchase probabilities were less than 0.50.

Wilks' lambda, the test of the null hypothesis that the group centroids occupy the same positions for the two groups, was found to be significant at the 0.005 level. This significance test was performed by a multivariate F-ratio, 2.2999 with 20 and 197 degrees of freedom. This suggested that the centroids of the two groups occupied different positions in the one-dimensional discriminant space. The group centroid was 0.7821 for intenders and -0.1099 for nonintenders.

Inspection of both the mean vectors and univariate F tests in Table 1 indicated that the variable innovativeness (#13) was the single most significant variable (P<0.025) that discriminated between the intender and nonintender groups for living room furniture buying intentions. It was also shown that nonintenders had a smaller mean vector than intenders. It was observed that husband intenders were generally high risk takers, had a better occupation and were less price conscious in their shopping attitudes. However, a noticeable difference was observed in the family size and level of education of intenders. The intenders' level of education was somewhat higher and their family sizes were smaller than nonintenders.

Examination of the scaled vectors of discriminant weights in Table 1 revealed that innovativeness made the greatest positive contribution to the centroid configuration, and to a lesser degree, education, a social factor, occupation, and bargaining. The highest negative coefficients were associated with low price option, a product durability factor, age, channel environment factor, and family size. The items above closely corresponded to the items that were significant in the univariate F tests. The variables and their discriminant coefficients seemed to form a kind of "innovativeness-price dimension," but the meaning of the dimension was not clear.

A multiple discriminant analysis as a procedure to classify individuals into the intender and nonintender groups, was also performed. The classification results were:




The overall accuracy of classification showed that 144 subjects, or 66%, were correctly placed out of 218. Because the maximum chance criterion was 57.8%, 66% classification was not a good fit when considering the fact that the same subjects were used for classification; hence, overestimating the fit.

The proportional chance criterion in this analysis was 51.25. However, given the chance criterion of 51.2%, 55.5% were classified in the nonintender group; the outcome should have been .509, or about 51% correctly classified. The classification of individuals for the nonintender group was greater than the chance criterion; although the difference was still small,of the 97 individuals classified as intenders, 60 were correct. This was 61.9% compared with a chance percentage of 42.2%.

Discriminant Analysis For Wives: For the analysis of differences between intenders and nonintenders for wives, 89 wives were classified as intenders and 133 were as nonintenders out of 222 wives. Table 2 provides the results for the intender and nonintender groups.

Wilks' lambda was found not to be significant for these two groups. The multivariate F-ratio was 1.1033 with 20 and 201 degrees of freedom. This means that the centroids of the two groups occupy similar positions (i.e., group centroid for intenders was -2.0611 and -2.5755 for nonintenders) in the one-dimensional discriminant space.

Only one variable, type of housing t#2), was significant at the 0.05 level as was shown by the univariate F test. Examinations of mean vectors for the two groups indicated that the intender group was at the low end of the scale, meaning that more intenders were residing in the duplexes or houses which may not be furnished or partially furnished. Other characteristics of intenders were as follows: They were slightly younger, had smaller family size, earned slightly higher income, were higher risk takers, and were less price-conscious than wife nonintenders.

Inspection of the scaled discriminant vectors of discriminant weights in Table 2 showed that the social factor (#19) made the highest positive contribution to the centroid configuration although the group separation was not significant in a statistical sense. Given the 20 variables in the analysis, the social factor had the greatest mean difference; and it was demonstrated that nonintenders were more concerned with the social acceptance and the conspicuousness of the product in expressing a buying intention. The highest negative coefficients were associated with type of housing (#2), family size (#7), low risk (#10), and economic shopping (#11). The variable and their discriminant coefficients seemed to form a kind of "social-socioeconomic dimension," but again interpretation was difficult because the groups were not different in a statistical sense.

The classification of individuals by the function into groups-was performed with the following results:




The proportional chance criterion was 51.96%, and the maximum chance criterion was 59.9%. In this analysis, 132 subjects or 59.5% were correctly classified out of 222. Because the maximum chance criterion was 59.9%, the 59.5% correct classification was a poor fit. Given the proportional chance criterion of 52%, 51% were classified as intenders and 49% as nonintenders. Both were poor classifications. Of the 113 subjects classified as intenders, 56 were correct. This was 51% to be compared with a chance percentage of 40%.


The results of discriminant analysis for husbands and wives demonstrated some interesting phenomena. Before comparing the results from the two different sex groups, there are several points worth mentioning about discriminant analysis and the classification procedure. First, close inspection of Table 2, for example, demonstrated that the variables that were significant under the univariate F tests and the scaled vectors that had either the highest positive coefficient or the highest negative coefficient did not necessarily contribute largely to the group centroid configuration. It was rather deceiving. The reason was that in this analysis, scale units of the variates were of several different types; scale units were not the same for all 20 variates. Variables from 9 to 16 (i.e., attitudinal classification scores), for example, could assume the values from 1 to 5, and buying intentions scores (variables 17 through 20) could assume the values from 1 to 7; while present income could take on any value from 1 to 100, or age could assume any values from 1 to 30. The discriminant weights, on the other hand, were independent of the units of measurement and the origin of coordinates of the original variates because the coefficients automatically adjusted themselves to the scales employed. Therefore, the problem occurs when the weights were multiplied by the means, and these products were summed across variates--the sum (centroid) was affected not only by the value of the discriminant vector but also by the scale units of the particular variable.

In view of the above discussion, it could be seen that discrimination on the variable, a social factor, was not really high for the wife groups because the scale units of the variate were small relative to the income variate. Thus, the variable making the highest positive contribution to the centroid configuration was the low-price option, and to a lesser but still significant degree, place of residence, income, and a social factor.

Secondly, the twenty variables used in this analysis contributed a great deal to the centroid configuration for the husband groups. Also the classification of husbands into these groups was fair. However, this was not true for the wife groups. There was no statistical significance in discriminating the intender and nonintender groups for wives, and the classification was poor. It should be noted that multiple discriminant analysis classified only on the variables given in the analysis. Poor discrimination between groups could be due to the fact that the most important variables for a particular group of people were overlooked, and hence were not included in the study. Therefore, it may be necessary in the future to select more relevant variables, and it should be possible to assess costs of misclassification of individuals into the designated groups.

Results from the discriminant analysis demonstrated noticeable differences between the groups in expressing buying intentions. With the lack of actual purchase data, it was difficult to assess as to whose buying intentions more accurately described the family's future buying plans. At the moment these study results appeared to support the argument that it is necessary to obtain two measures of buying intention, one from the husband and the other from the wife, from each household rather than relying on a single measure of buying intentions. These two separate measures of buying intentions may help to improve the prediction of the actual purchases.

The differences between husbands and wives could be further supported by examining the joint distribution of the rank order of the purchase priorities on living room furniture given by husbands and wives. Table 3 describes the rank order of each of the sixteen products (Suh, 1972a) from the first choice to the fifth choice. Because the number of subjects whose rank order was less than six was negligible, those were combined. Diagonal entries indicate the number of couples who were in agreement in their preferences to purchase a given product.



It was apparent for living room furniture that (a) the aggregate of the rank order between husbands and wives was not a perfect match, nor was it a poor one, which was indicated in each of the column and row totals. (b) It was even more apparent that there was very poor agreement within a family given that each of the husbands and wives expressed their buying intentions. For example, of the 27 wives (see row total) and 21 husbands (see column total) who gave first priority to the purchase of living room furniture, only eleven couples from the same families were in agreement. (c) Finally, agreements became poorer as the rank order decreased, except in the case when couples were expressing no-buying-intentions (i.e., zero rank order).

It appeared that the poor agreement shown in the joint distribution had a carry-over effect in the discriminant analysis. As it had been seen in the two different analyses, agreement amongst the variables making contribution to the centroid configuration was poor, and even the subsequent discriminant coefficients for variables were different from one to another. Hence, the study raises serious doubts on consumer surveys which simply rely on a single response from each family. This appears to be more difficult and problematic than the previous findings have reported. Differences in expressing buying intentions between husbands and wives do not simply guarantee the reliability of information obtained from a member of the family.


Blood, R. O. & D. M. Wolfe. Husbands and Wives: The Dynamics of Married Living, Glencoe, Illinois: The Free Press, 1960.

Byrnes, J. C. An Experiment in the Measurement of Consumer Intentions to Purchase. Journal of American Statistical Association: Business and Economics Statistical Section, 1964, 59, 265-279.

Davis, H. L. Dimensions of Marital Roles in Consumer Decision Making. Journal of Marketing Research, 1970, 7, 168-177.

Ferber, R. On the Reliability of Purchase Influence Studies. Journal of Marketing, 1955a, 19, 225-232.

Ferber, R. On the Reliability of Responses Secured in Sample Surveys. Journal of American Statistical Association, 1955b, 50, 788-810.

Ferber, R. & R. A. Piskie. Subjective Probabilities and Buying Intentions. Review of Economics and Statistics, 1965, 47, 322-325.

Granbois, D. H., and R. P. Willett. Equivalence of Family Role Measures Based on Husband and Wife Data. Journal of Marriage and The Family, 1970, 32, 68-72.

Haberman, P. W. & J. Elinson. Family Income Reported in Surveys: Husbands versus Wives. Journal of Marketing Research, 1967, 4, 191-194.

Heer, D. M. Husband and Wife Perceptions of Family Power Structure. Marriage and Family Living, 1962, 22, 65-67.

Heer, D. M. The Measurement and Bases of Family Power: An Overview. Marriage and Family Living, 1963, 23, 133-139.

Juster, F. T. Anticipation and Purchases: An Analysis of Consumer Behavior. New Jersey: Princeton University Press for National Bureau of Economic Research, 1964.

Juster, F. T. Consumer Buying Intentions and Purchase Probability: An Experiment in Survey Design. Occasional Paper 99. New York: National Bureau of Economic Research, 1966.

Namias, J. Intentions to Purchase Compared with Actual Purchase of Household Durables. Journal of Marketing, 1959, 24, 26-30.

Namias, J. Intentions to Purchase Related to Consumer Characteristics. Journal of Marketing, 1960, 25, 32-36.

Piskie, R. A. A Quantitative Measure of Consumer Buying Intentions. Unpublished Masters Thesis, Department of Economics, University of Illinois, Urbana, Champaign, Illinois, 1963.

Sharp, H. and P. Mott. Consumer Decisions in the Metropolitan Family. Journal of Marketing, 1956, 21, 149-156.

Suh, R. H. An Application of Multivariate Statistical Techniques for the Study of Consumer Buying Intentions for Durable Goods. Unpublished Doctoral Dissertation, College of Commerce and Business Administration, University of Illinois, February, 1972.

Suh, R. H. Response Patterns of Durable Goods Buying Intentions by Husbands and Wives--A Methodological Question. (In process), 1972.

Wallin, P. and A. Clark, Cultural Norms and Husbands' and Wives' Reports of Their Marital Partners' Preferred Frequency of Coitus Relative to their Own. Sociometry, 1958, 21, 247-254.



Raymond H. Suh, California State University, Long Beach


SV - Proceedings of the Third Annual Conference of the Association for Consumer Research | 1972

Share Proceeding

Featured papers

See More


O1. Choice, Rejection, and Context Effects

Shih-Chieh Chuang, National Chung Cheng University
Yin-Hui Cheng, National Taichung University of Education

Read More


L5. Understanding the components and effects of the Omnichannel Seamless Experience.

PAULA RODRÍGUEZ-TORRICO, Universidad de Burgos (Spain)
Lauren Trabold, Manhattan College
Sonia San-Martín, University of Burgos (Spain)
Rebeca San José, University of Valladolid (Spain)

Read More


From a Culinary Phantasm to an Eudaimonic Well-Being : Exploring The Experience of Amateur Cooking Classes

Virginie Brégeon de Saint-Quentin, Ferrandi Paris, the french school of gastronomy
Ophélie Mugel, Chaire SDSC AgroParisTech, Université Paris Est

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.