Household Durable Goods Acquisition Behavior: a Longitudinal Study

ABSTRACT - The concept of "priority" or sequence of acquisition reflecting the order in which households acquire items has been one of the dominant paradigms employed for understanding durable goods consumption and purchase decision behavior. So far, studies have only examined the issue descriptively using crosssectional ownership data. The present study tests a theoretical and predictive model of durable goods acquisition behavior based upon thirteen years of decision, consumption, and ownership behavior for the same households. The results indicate that the proposed model has strong explanatory and predictive capabilities.


Michael C. Mayo and William J. Qualls (1987) ,"Household Durable Goods Acquisition Behavior: a Longitudinal Study", in NA - Advances in Consumer Research Volume 14, eds. Melanie Wallendorf and Paul Anderson, Provo, UT : Association for Consumer Research, Pages: 463-467.

Advances in Consumer Research Volume 14, 1987      Pages 463-467


Michael C. Mayo, The University of Michigan

William J. Qualls, The University of Michigan


The concept of "priority" or sequence of acquisition reflecting the order in which households acquire items has been one of the dominant paradigms employed for understanding durable goods consumption and purchase decision behavior. So far, studies have only examined the issue descriptively using crosssectional ownership data. The present study tests a theoretical and predictive model of durable goods acquisition behavior based upon thirteen years of decision, consumption, and ownership behavior for the same households. The results indicate that the proposed model has strong explanatory and predictive capabilities.


The acquisition and consumption of household durable goods continues to be a topic of interest among marketers and researchers. A review of the literature indicates the perspectives of economists (durable good forecasts), sociologists (household decision behavior), and psychologists (individual consumer behavior) have contributed to the body of knowledge regarding household durable goods behavior. Yet, as evidenced by the recent research grants awarded by Marketing Science Institute, issues regarding the consumption of household durables remain unanswered.

The current research in this area by consumer researchers can be classified into two types: 1) Descriptive analyses of planning processes and household types (Cox, Granbois, and Summers 1983, Corfman and Lehmann 1985) and 2) determination of acquisition and consumption patterns of household durable goods (Kasulis, Lusch, and Stafford 1979, Dickson, Lusch, and Wilkie 1983). While past research has produced useful conceptual frameworks for addressing the issues discussed above, the methodologies employed have suffered from several serious limitations. For example, the shortcomings of the Guttman Scalogram analysis technique, employed in both the Kasulis et al. (1979) and Dickson et al. (1983) studies have been clearly addressed in the research literature (Clogg and Sawyer 1981). In addition, research to date has attempted to model household durable goods behavior as a static point in time based upon cross-sectional data (Cox, et al. 1983, Corfman and Lehmann 1985) instead of examining the phenomena for the same household from a longitudinal perspective.

The present study attempts to address four critical issues regarding the acquisition and consumption of durable goods. Specifically, the paper presents the results of an investigation which:

1. Examines the underlying structure of durable goods acquisition and consumption patterns of the same households over a 13-year period.

2. Develops a classification of household types based on their durable goods acquisition and consumption behavior.

3. Tests a theoretical framework of why households acquire and consume durable goods in the manner in which they do.

4. Tests a model for predicting future acquisition of household durable goods.

Two modeling techniques, latent structure analysis (LSA) and partial least squares (PLS), are applied to data from the Illinois panel study of consumer decision processes to describe and predict the acquisition and consumption of household durable goods. The next section provides a brief critique of the research in this area. Next we discuss the sample and the two analytical techniques employed in the study. Finally results are presented and discussed and suggestions for future research are offered.


One outcome of research in this area has been the classification of households based on their durable goods ownership and/or purchase plans. Alderson (1957) refers to the process as household durable goods assortment management. It is based on the premise that all newly formed households start out with a "starting set" of durable goods acquired through gifts, purchases, lease/ rentals, previous ownership, or as part of the initial home dwelling. Future durable goods acquisition and consumption becomes a function of this "starting set."

Newly formed households are seldom able to purchase the complete set of durable goods necessary to fully furnish and stock a household. Thus families (primarily husbands and wives) must decide on an order of purchase, and a decision plan regarding how these purchases will be made over time. The idea of an ordered sequence or priority pattern reflecting the process by which households acquire durable goods has received considerable attention from researchers (McFall 1969, Lusch, Stafford, and Kasulis 1978, Kasulis, Lusch, and Stafford 1979, Dickson, Lusch, and Wilkie 1983). Evidence from these studies supports the contention and demonstrates the existence of some underlying priority pattern or order in which household durable goods are purchased.

A study reported by McFall (1969) found an acquisition priority pattern for a set of "comfort" products (electric blanket, washing machine, room air conditioner, and dishwasher) from which it was concluded that a household which owns a dishwasher also owned the remaining three durable goods. A similar acquisition pattern of household "comfort" durable goods was found by Lusch et al. (1978) in a cross-sectional survey of over 1,800 households (washer, dryer, dishwasher, freezer, and microwave ovens). Both studies used the actual ownership of durables as the basis for determining acquisition patterns. While such studies established that there are indeed a common set of durable goods that are purchased by households, the prediction of future acquisition behavior is poor due to the lack of consideration of future purchase priorities or changes in household circumstances.

The relationship between planned purchases and actual purchase behavior has always been tenuous at best. Dickson and Wilkie (1978) found that there were a large number of unfulfilled durable purchase plans (and purchases made with no plans) when compared to the households reporting purchase plans. One reason put forth by researchers to explain the lack of correlation between planned and unplanned purchase of durable goods has been changes in household circumstances. Pickering (1975) found that multiple reasons were often given, but household financial problems and changes in priorities were the most frequent reason given for households not following through with durable goods purchase intentions.

Beyond the "starting set," Granbois (1977) suggests that four reasons account for household durable good acquisition: 1) Maintenance Replacement, 2) Adjustment/Upgrading Replacement, 3) Additional Unit Expansion and 4) First Acquisition Expansion. More recently Cox et al. (1983) found that while such acquisition categories partially explain the variation in household durable purchase behavior (search, satisfaction, and payment method) the researchers were skeptical about interpreting its generalizability beyond the reported study.

While the brief review provided here only touches the surface, the findings are indicative of the state of the art regarding household durable goods acquisition and consumption behavior. For a detailed review the interested reader is referred to Dickson and Wilkie (1978). Our study is a step toward identifying-the underlying structure of durable goods acquisition behavior with methods which may greatly improve our ability to understand this consumption process.



The data for the present study were collected by the Survey Research Lab at the University of Illinois as part of a much larger study of household consumer decision behavior. The database was constructed from panel data obtained from an initial sample of 311 household couples who were married in Peoria or Decatur, Illinois during the summer of 1968. Eighteen interviews were conducted at regular intervals from the Fall of 1968 through the Summer of 1981.

The sample used in the present study controlled for three key variables to allow for the tracking of the same households over time (13 years). Specifically, only married couple households who stayed together for all eighteen waves of the panel study, who rented their living quarters, and who owned only one car were selected for the present analysis. Unfortunately, such controls which help to increase the integrity of the data, cost in terms of a reduction in sample size. Thus an initial sample size of 311 was reduced to 146 which decreased to 61 households by the time of the last interview. Yet, during this time period over 1600 durable good purchases were made.


The data available for this study are particularly conducive to the examination of proposed theoretical relationships of durable goods acquisition and consumption behavior. The same households have been traced over 18 waves of data collection. For the purposes of this study, data have been aggregated over time to reflect the stages of the household life cycle (Bristor and Qualls 1984). Specifically, waves 1-5 were treated as newly formed households, waves 6-9 as early adaptive households, and waves 10-14 as late adaptive households. Waves 15-18 (mature households) were intentionally held out for predictive purposes.

The set of durable goods examined included a diverse set of goods which are typically found in a household. Ten (10) household durables composed the set of goods examined over the life cycle of the study. Specifically, they included first car, washing machine, stove, refrigerator, dryer, color TV, stereo, second car, freezer, and dishwasher.

The dependent variable for the PLS models is future durable ownership. The independent variables include prior ownership and likelihood to buy (LTB). Prior ownership refers to the set of durable goods held by a household at a given point in time. LTB is defined as the probability of purchasing a given durable good at some future time period. Only households with a 75 percent chance or better of purchasing some durable goods were included in the actual analysis.

Method of Analysis

In the present study, two structural modeling techniques are employed to analyze the durables' purchase data to investigate their efficacy in describing acquisition patterns and predicting future household purchase behavior. The two techniques employed allow the use of categorical data, which is typical for most durable goods acquisition and consumption studies. Latent structure analysis and partial least squares are used to analyze the data.

Latent Structure Analysis

Latent structure analysis (LSA) is a technique used to describe a set of models whose purpose is to mathematically assign observations to one or more unobserved (latent) classes. The objective of LSA is to characterize latent variables that explain the observed relationship between theoretical constructs. LSA evolved from Guttman's early work and was conceptualized by Lazarsfeld in the 19508 and further developed by Lazarsfeld and Henry (1968).

A latent class model can best be described as a "data unmixing procedure" which starts with data in the form of a multiple contingency table that determines the association between the categorical variables making it up. The technique assumes that there are a certain number of manifest (observable) variables (i.e., A,B,C) each with a given number of levels (i.e., i = 1, 2....I, j = 1, 2....J, k = 1, 2....K). As such the relationship among A, B, C can best be explained by a single latent factor reflected by X with T classes; based upon the probability that an individual is in a specific cell of the contingency table. Specifically:


Lazarsfeld and Henry (1968) suggest that "Within a latent class T, responses to different items are independent. The within class probability of any pattern of response to any set of items is the product of the appropriate marginal probabilities." Once the probabilities of class membership in A, B, C and X, have been estimated, the fit between the observed proportion (R) and estimated proportion (E) of a specific model can be assessed via two different test statistics: (1) Pearson Goodness of fit Chi square (2) A Chi-square based upon the likelihood ratio criterion.

The application of LSA techniques to marketing problems is not new as evidenced by the literature, (Green, Carmone, and Wachspress 1976, Dillon, Madden, Mulani 1983, Clogg, and Munch and Callahan 1983). The basic purpose of using LSA in the present study was to determine whether or not an underlying structure for durable goods acquisition classes over a period of 13 years could be identified. The computer program Maximum Likelihood Latent Structure Analysis (MLLSA) was used for the analysis.

Partial Least Squares

One of the objectives of partial least squares (PLS) is to maximize the explainable variance of both latent and manifest variables. PLS is a predictive structural modeling technique which estimates a model's parameters through a series of ordinary least squares regressions, deriving its predictive power by minimizing residual variances. Unobservable variables are estimated as exact linear combinations of their empirical indicators

h = Phy and x = Pxx

where En and EX are regression matrices. The model is estimated via a set of weights/loadings which describe the relationship between unobservable and observable variables. Unlike the more popular structural modeling program: Lisrel, PLS is more robust regarding assumptions about the population, measurement scales and distribution, (Wold 1980).

Lohmoller's LVPLS 1.6 computer program was used to generate the analysis. PLS as a structural modeling technique is beginning to receive considerable attention in the marketing literature (Fornell and Bookstein 1982, Fornell and Larcker 1981, Fornell and Robinson 1983). Recent work by Wold and Bertholet (1985a, 1986b) has shown that model building techniques using PLS estimation procedures are applicable when analyzing multidimensional contingency table analysis.


LSA is used to uncover the underlying structural classes of household durable goods and acquisition behavior. The LSA model was run as 1, 2, and 3, latent hypothesized classes and tested on each of the first three waves separately and on the panel as a whole. In comparing the 1, 2, and 3 class models, we fount that the three class model fit best in terms of a maximum likelihood Chi-Square (p s .99), a Pearsonean Chi-Square (p - .99), and an index of dissimilarity (.0002) generated by the MLLSA program. ese three indices indicate that the three class model is consistent with the data.

Since we are interested in the underlying structure of durable acquisition behavior, and the amount of each set of durables within each latent class, one can reconstruct a contingency table from the final conditional probabilities, generated in the MLLSA program. The set of durable goods were grouped into four categories based upon the three latent class model of waves 2-14. the categories are:

1) Primary transportation: Replacement of first car given one and only one car owned at the beginning of the study (represents the sample control category).

2) Basic household: includes the stove, refrigerator, washer, and dryer.

3) Standard luxury: includes second car, color TV, and stereo.

4) Luxury comfort: includes the dishwasher and freezer.

LSA derived conditional probabilities for the items within the 3 latent classes. While several predictions are close, most are completely inaccurate. Table 1 summarizes these findings. In attempting to explain the lack of convergence between actual and predicted durable purchases, one can look at the way the final conditional probabilities are calculated. The researcher must make initial estimates of the conditional probabilities based on a hypothesized theory regarding the nature of the underlying latent classes. Unless these initial estimates are close to the real estimates, there will be a large variance between estimates and actual behavior. While the MLLSA program will usually converge to a solution, the longer it takes to converge, the less accurate the estimated probabilities will be. The results from the present analysis converged after 45 interactions which is relatively few but still generated inaccurate conditional probabilities.



While the identified latent classes are not intended to be exhaustive of all possible categories, it represents a reasonable way to classify households over its life cycle (HLC). Based upon the classifications, the data were then subjected to PLS analysis testing the latent class model uncovered by LSA.

PLS Analysis of the Three Latent Class model

To test the existence and impact of the three latent class or household type model, the model presented in Figure 1 was analyzed via PLS. As the model illustrates, there are four manifest variables for the first latent variable: likelihood to buy (e1), corresponding to the four durable categories mentioned earlier. The manifest variables of the second latent variable (e2) represent the set of durable goods acquired by newly formed households during waves 2-5 (a period from 7/69 to 2/71). Measures of likelihood to buy were obtained from each household for each durable product and averaged across categories within household types. It is hypothesized in the present model that the likelihood to buy and present ownership of durables is a strong indicator of future household durable good acquisition and consumption behavior. Over the life cycle of the household as ownership and acquisition patterns change, the likelihood of buying specific household items should change, and vice-versa. us there should be a positive relationship between likelihood to buy and durable good purchase behavior. Figure 1 summarizes the results of the PLS analysis and indicates that in the earliest stage of the HLC, there is no relationship between LTB and purchase behavior. In the later part of the HLC, LTB is positively related to acquisitions and increases in influence from the early adaptive to the late adaptive stages. Acquisitions in the newly formed stage have a strong, positive influence on acquisitions in the early adaptive stage but, although positive, the strength diminishes in the later stages. e explained variance captured by (n1). (n2). and (n3) is encouraging.

In examining the loadings on the indicators of the Acquisition Latent Variables, one finds that Standard Luxury items represent the acquisition pattern of the newly formed household. is is not unexpected in that all families started with one car and it is unlikely that it was replaced early in the HLC. Further, since these couples are renters, they probably had access to the basic household items. In the early adaptive stage, the acquisition of Standard Luxury items decreases while couples begin to acquire the Basic Household items. The replacement of the first car increases from -.10 to .06 indicating some replacement activity. Five years had elapsed from the beginning of the survey to the end of the early adaptive stage so this behavior is not unexpected. In the late adaptive stage, the acquisition of Basic Household items continues and Luxury Comfort items become prominent. The increase of the Luxury Comfort items in this late stage is a reasonable expectation.




In general, the PLS analysis provides the researcher with some useful insights regarding household durable goods acquisition. Early models of durable goods acquisition behavior were never able to detect such changes in priorities. Such changes should have a big impact upon predictions of future durable goods acquisition behavior and, therefore, should be explicitly modeled into durable good forecasts. This issue is addressed in the next section.

A Predictive Model of Household Durable Goods Acquisition

A third objective of the present paper has been to develop a method for predicting household durable good acquisition behavior. Using the last three years of ownership data obtained from the household panel data base for validation, a predictive model of what durables would be acquired during the next stage of the HLC was developed and tested. Thus a model (Figure 2) was constructed in an attempt to predict durable purchases during waves 15-18. Again the two major determinant variables hypothesized in the model include, 1) likelihood to buy (t1) and 2) durable ownership (i2) (as of wave 14). Since the actual purchases of the last three years had been held out, the present study is able to compare the predictive purchases with the actual purchases of durable goods by the households. The results of the predictive test are presented in Figure 2 and Table 2 respectively.

The path analysis has the signs that one would expect. Household durables which people now own are negatively correlated with what is to be purchased during the next three years. Conversely, the likelihood to buy is positively correlated with future purchase behavior. The comParative results of the model are presented in Table 2. As can be seen, predictions for primary transportation and standard luxury durable goods are quite good, while the prediction of basic household and luxury comfort goods are poor.







The reported path coefficient (.32) is reasonable for the likelihood to buy link to future ownership. Support for the findings of earlier studies which suggested a poor relationship between planned purchases and actual purchases was not found. It should be remembered that only households which reported a 75 percent or greater probability of purchase of a certain set of durable goods were used in the analysis. As a result, sparse LTB data may destabilize PLS estimations. Conversely, the poor explanatory power of the hypothesized model (R2 - .15) suggests that there are other factors which come into play which affect the household durable purchase and acquisition process.


The purpose of this paper has been to expand upon earlier research on household durable goods acquisition and consumption behavior. Specifically, a model was developed and tested which explains and predicts durable goods acquisition behavior. The present research has led to what may be classified as a theoretical framework of durable goods acquisition and development of a base from which to conduct future research.

The present study is an improvement over earlier studies in three ways. First the durable decision and purchase behavior of the same households are tracked and analyzed over a thirteen-year span; previous studies only examined the issues cross-sectionally at specific points in time (2 years). Second, the study is one of the first attempts to develop and test a theoretical framework of durable goods acquisition behavior. Both behavioral (likelihood to buy) and structural (position in HLC) influences are modeled to explain variations in household behavior. Finally a predictive model based on a theoretical framework is tested and partially validated.

Two analytical techniques (LSA and PLS) were employed to improve upon the methodologies used in earlier studies. LSA proved to be a useful technique for determining the underlying durable acquisition structure for households. Three latent classes appear to be a reasonable representation of the underlying structure of the data. Despite the problems with the final conditional probabilities of cell membership, the use of LSA in this application is an improvement over similar studies using Guttman Scalogram Analysis.

The results from the PLS analysis also suggest an order of acquisition with respect to household durables. During the newly formed HLC stage, households can be characterized by an ownership preference for basic household items and luxury entertainment goods. Conversely, by the end of the first stage of the HLC, luxury comfort items were not heavily stocked by households (despite the high loading) and there is very little emphasis on the replacement of the family's first car. During the adaptive stage of the HLC a shift from luxury entertainment and luxury comfort goods to the replacement of the first car is suggested. The acquisition pattern uncovered via PLS is a better predictor than previous models given the addition of other explanatory variables (i.e., LTB). Yet, the discrepancies regarding the direction and strength of the hypothesized relationship should temper the explanatory power exhibited by the model.

The managerial usefulness of identifying durable goods acquisition patterns is well documented. Specifically, marketers interested in forecasting consumer demand and/ or targeting market segments would be interested in such results. As such the predictive model tested should be of value to marketers. Although the explanatory power of the model is low, the hypothesized relationships generally carried the expected signs and are in the right direction. This would suggest that there are additional influences not accounted for in the hypothesized model that should be investigated in future research.

Theoretically, future research could address the issue of refining the model by adding additional explanatory variables and retesting the model. For example, household factors such as role structure, occupation, and income could aid in increasing the explanatory power of the model.

Finally, the two analytical techniques (LSA and PLS) appear to be complimentary in their relationship. Together they appear to offer researchers insight and improvements, not found in other methodologies. Thus the study of acquisition patterns coupled with the LSA and PLS methodologies can be a fruitful avenue for marketing researchers and practitioners.


Bertholet, J. L., and H. Wold, (1984), "Recent Developments on Categorical Data Analysis by PLS Modeling. In ICUS Seminar.

Bertholet, J. L., (1977), "Unrestricted and Restricted Maximum Likelihood Latent Structure Analysis: A Manual for Users," Population Issues Research Office, The Pennsylvania State University, Working Paper #1977-09.

Bertholet, J. L., J. M. Munch and D. G. Callahan, (1983), "Application of Latent Structure Models in Marketing Research: Exploratory Analysis of Buying Style Items," Advances in Consumer Research, Vol. X, R. Bagozzi and A. Tybout (eds.) Association Consumer Research. 32-36.

Dickson, P. R., R. Lusch, and W. Wilkie, (1983), "Consumer Acquisition Priorities for Home Appliances: A Replication and Re-evaluation," Journal of Consumer Research, Vol. 9, (No. 4), 432-435.

Dickson, P. R., and W. Wilkie, (1978), The Consumption of Household Durables: A Behavioral Review, Marketing Science Institute. Cambridge, Mass.

Kasulis, J., R. Lusch, and E. Stafford, (1979), Consumer Acquisition Patterns for Durable Goods,' Journal of Consumer Research, Vol. 6, 47-57.

Lazarsfeld, P. and N. Henry, (1968), Latent Structure Analysis, Boston: Houghton Mifflin.

Lohmoller, J. B., (1981), LVPLS 1.6: Latent Variables Path AnalYsis with PLS Estimation: Program Manual, Forschungsbericht 81.04, Hochschoule der Bundeswehr, Munchen.

Lusch, R. F., E. F. Stafford, and J. J. Kasulis, (1978), "Durable Accumulation: An Examination of Priority Patterns, Advances in Consumer Research, Vol. 5, H. Keith Hunt (ed.) Association for Consumer Research, 119-125.

Lusch, R. F., and L. Bertholet, (1985b), "Recent Developments in Categorical Data Analysis by PLS," Measuring the Unmeasurable, P. Nijkamp, H. Leitner, N. Wrigley (eds.) Martinus Nithoff Publishers, Boston, 253-286.

(A complete list of references are available upon request from the authors.)



Michael C. Mayo, The University of Michigan
William J. Qualls, The University of Michigan


NA - Advances in Consumer Research Volume 14 | 1987

Share Proceeding

Featured papers

See More


L12. Should I Stay or Should I Go: When Our Companies Have Eyes for Other Consumers

Na Ri Yoon, Indiana University, USA
Jenny Olson, Indiana University, USA
Adam Duhachek, Indiana University, USA

Read More


Why is 1 out of 20 Riskier than 5%? Effect of Representing Unlikely Events as Frequency versus Percentage on Risk Perceptions

Nevena T Koukova, Lehigh University
Joydeep Srivastava, Temple University, USA

Read More


Expressing Dissent: How Communication Medium Shapes Dehumanization and Attitude Change

Juliana Schroeder, University of California Berkeley, USA

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.