Who Is the Deal Prone Consumer?

ABSTRACT - This paper uses a model of consumer buying behavior to identify household characteristics that should affect deal proneness. The model assumes that the household makes purchasing and inventory decisions much like a firm. In other words, the household's purchasing decisions are based on factors such as transaction costs, holding costs, and stockout costs in addition to product price. Household characteristics are then related to these cost parameters to predict which households are likely to be deal prone. The predictions are tested empirically using panel data on five frequently-purchased products. The empirical results indicate that it is possible to identify the deal prone household and that the key variables that affect deal proneness are household resource variables such as home ownership and automobile ownership.


Robert Blattberg, Thomas Buesing, Peter Peacock, and Subrata Sen (1978) ,"Who Is the Deal Prone Consumer?", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 57-62.

Advances in Consumer Research Volume 5, 1978      Pages 57-62


Robert Blattberg, University of Chicago

Thomas Buesing (student), University of Chicago

Peter Peacock, Wake Forest University

Subrata Sen, University of Rochester

[This research has been funded, in part, by National Science Foundation Grant SOC73-05547.]


This paper uses a model of consumer buying behavior to identify household characteristics that should affect deal proneness. The model assumes that the household makes purchasing and inventory decisions much like a firm. In other words, the household's purchasing decisions are based on factors such as transaction costs, holding costs, and stockout costs in addition to product price. Household characteristics are then related to these cost parameters to predict which households are likely to be deal prone. The predictions are tested empirically using panel data on five frequently-purchased products. The empirical results indicate that it is possible to identify the deal prone household and that the key variables that affect deal proneness are household resource variables such as home ownership and automobile ownership.


Marketing managers have always been interested in identifying the deal-prone household on the basis of available demographic data. If such households can be identified precisely, specific marketing strategies designed to appeal to such households are likely to be more effective. For example, demographic information is available by zip codes or census tracts, if certain demographic groups are more deal prone, coupon distribution could be restricted to those areas where households with higher deal proneness reside. This would reduce couponing costs with a less than proportionate reduction in response. Similarly, more accurate identification of deal prone households would increase the marketer's ability to match deal prone households and media audience characteristics, thus increasing the efficiency of media distribution of coupons and other promotional items.

Several studies have tried to identify the deal prone household. Webster (1965) and Montgomery (1971) have published two of the better known studies. The results of these and other studies are summarized by Frank, Massy, and Wind (1972, p. 124) who state:

The results of cross-sectional studies, almost without exception, indicate that there is, at best, only a modest degree of association between demographic, socioeconomic, and/or personality characteristics, and selected aspects of household purchasing behavior, such as total consumption, brand loyalty, and deal proneness.

One reason for this "modest degree of association" may lie in the methodological approach usually taken in these studies. Typically, a large number of potential explanatory variables are regressed against the proportion of purchases made on deal in a search for statistical significance. For example, Webster (1965) ran 200 regressions with different combinations of 45 explanatory variables. This approach is open to serious question because one cannot always determine if "significant" relationships reflect a valid relationship or a spurious one which has arisen by chance alone. Without a theory to indicate which variables should affect deal proneness, there is a greater danger of accepting spurious results.

A related deficiency of prior studies of deal proneness arising from the absence of a clearly stated theory has been improper specification of explanatory variables. For example, Montgomery (1971) included "presence of children" as an independent variable in a model but was unable to predict a priori whether the presence of children should or should not increase deal proneness. However, if one hypothesizes that it is the age of the children that affects deal proneness (rather than their presence per se), it is possible to predict a priori the impact of "age of children" upon deal proneness. If the children are below the age of six (and are consequently not yet in school), they require more of their parents' time, thus reducing the time available for shopping. Having less time to shop leads to fewer shopping trips and fewer opportunities to take advantage of deals, thereby reducing the household's deal proneness. Montgomery's results led him to conclude that presence of children was not related to deal proneness. However, by considering the presence of children without determining how their presence should affect deal proneness, Montgomery may have arrived at an incorrect conclusion about a variable which, if properly specified, could well be related to deal proneness.

Like the studies cited above, this paper also attempts to identify the deal prone household. However, the approach used here is quite different from that used in most of the earlier work. We begin by formulating a model of household purchasing behavior. The model is then used to predict how certain demographic variables should affect deal proneness. Finally, an empirical evaluation of the predictions is made. The empirical results show that it is possible to identify the deal prone household using demographic variables and that the effect of these variables is substantial.


Model Assumptions

A household inventory model is developed in this section which is based on the assumption that the household is the same as any producing unit which needs to stock inventory and meet demand. This assumption follows from the notion of the household as a production unit which Becker (1965) and others have used to model consumer behavior in the economics literature. This modeling approach is used because it has proved to be very fruitful in several applied studies in economics (Michael and Becker, 1973) and also because it appears to be a potentially promising approach in marketing (Kunreuther, 1973). The inventory model proposed here also corresponds closely to models developed by management scientists to make better inventory decisions in more traditional production environments (Hillier and Lieberman, 1974, pp. 472-527).

The initial assumption in the model is that households make long-run decisions about whether to use a given product at all and the average number of units of the product to use per period. Decisions about product use and the average usage rate are determined exogenously by factors such as family size, family income, etc. It is also assumed that ratios of prices of the product in question and prices of substitutes and complements are constant during the period considered so that households need not evaluate their usage rate decision because of changes in relative price. The latter assumption is made in order to obtain a tractable model and because it permits us to concentrate on the purchase timing decision which is the basic focus of this paper.

Cost Structure of the Household

Four categories of cost affect household inventory decisions: (1) transaction cost, (2) storage cost, (3) stockout cost, and (4) the actual price of the item. Transaction cost is the opportunity cost of the time required to purchase an item once the consumer is actually in a store plus the opportunity cost of travel time required to get to and from the store where the purchase takes place. Transaction cost will vary across stores. And if a consumer has a "regular" or preferred store, one would expect the transaction cost of an item purchased there to be less than if a special trip is made to purchase the item at some other store. Storage cost represents interest on the capital required to maintain a given level of inventory plus the cost of the required space. Stockout cost reflects the foregone utility of not consuming an item which is not in stock at the time it is demanded. If the household can easily substitute other items in the event of stockout, or if it derives little utility from consuming the item, stockout cost should be low. Observed price per unit is the final component of cost. For purposes of the analysis, it is assumed that the observed price in a store is constant within any given period, e.g., a week. Prices may vary across stores and may change from period to period.

Mathematical Formulation of the Model

The household's purchase decision process is represented mathematically in Exhibit 1. The model shown requires the consumer to minimize expected costs over a finite time horizon which includes present as well as future periods. Thus, expectations about future prices and future demand affect the present period's decision. Note that though the household's average demand per period is known, the exact quantity demanded in each period is unknown at the start of the period. Thus, this quantity, dt is represented as a random variable in the model. Future price expectations are generated by probability distributions of the time between deals and the length of time a given deal is in effect. These distributions would be based on actual household experience with deal duration and time between deals in each store.

For the model shown in Exhibit 1, it is assumed implicitly that all brands in the product class yield the same utility to the consumer. Therefore, the consumer's objective is simply to minimize the "total cost" of buying the product. Again, this assumption is made only to obtain a tractable model. In order to consider the case where different brands offer different utilities, a framework which jointly considers utility maximization and purchase timing would have to be developed.




The inventory model described above provides a theoretical basis to identify demographic and household resource variables which should lead to deal proneness. To see this, consider Exhibit 2 which provides a diagrammatic representation of the manner in which the inventory model links demographic and resource variables with household deal proneness. Exhibit 2 indicates that the three inputs to the inventory model are: (1) the price distribution of the product across the stores, (2) the household usage rate for the product (which, in turn, is influenced by demographic variables such as family size) and (3) household cost parameters.



Deal Proneness and Household Cost Parameters

The first link of interest is the one that connects household cost parameters with deal proneness through the inventory model. What does this link represent? Essentially, it indicates how the cost structure facing a household determines whether or not the household will be deal prone. For example, if storage costs were low, one would expect households to stock up on a commodity when a deal is on. Alternatively, if the transaction cost of traveling to a non-preferred store were sufficiently high, one would not anticipate a consumer to buy on a deal offered by the non-preferred store.

Cost Parameters and Household Demographics and Resources

But what factors determine the household's cost structure? To answer this question, consider the link in Exhibit 2 which connects Demographic Variables (like income) to Household Cost Parameters through Household Resource Variables (such as housing and transportation). Note first that income is an important determinant of household resources, i.e., households with higher income are more likely to own homes (as opposed to being renters) and are also more likely to own one or more cars. These household resources, in turn, affect the cost parameters of the model. For example, home owners typically have more storage space available compared to apartment dwellers and hence should incur lower storage costs Similarly, car ownership makes transportation easier, thereby reducing the household's transaction costs.

It was noted earlier that low storage costs and low transaction costs both lead to deal proneness. Since low storage costs are associated with home ownership and low transaction costs with car ownership, it is possible to make specific predictions such as the following: home owners and car owners will tend to be more deal prone than apartment dwellers and households without cars.

One can see, therefore, that the inventory model's links with deal proneness and with household cost parameters makes it possible to (1) identify the relevant variables that should affect deal proneness and (2) predict the direction of their effect. The model is also useful in providing a quantitative evaluation of the relative impact of the various cost parameters on deal proneness (see Blattberg, et. al., 1977).


In this section, we state formally our predictions of how some specific household resource and demographic variables lead to deal proneness. Because of the length constraints on this paper, only two types of explanatory variables will be studied: (1) the household resource variables mentioned earlier, i.e., car ownership and home ownership, and (2) income. Note that the data for these variables are available by zip codes or census tracts. Therefore, if these variables do affect deal proneness, the marketing manager can implement the re-suits easily. This is in contrast to some of the variables found by Webster (1965) and Montgomery (1971) to affect deal proneness. For example, both found that brand loyalty was negatively associated with deal proneness. However, one must first identify who the less brand loyal consumers are before one can use such a finding.

Household Resource Variables and Deal Proneness

A key component of the transaction cost of shopping is transportation cost. Households that do not have cars available are forced to shop at stores that are nearby. They are also more likely to shop at a single store (see Kunreuther, 1973, p. 376). Since the ability to take advantage of deals depends upon the freedom to shop often and at many stores, households without cars should be less deal prone.

The second household resource variable is home ownership This variable should be related to holding costs. Apartment dwellers usually have less storage space available than homeowners simply because apartments are smaller. Therefore, holding costs should be higher for apartment dwellers. Since lower holding costs should lead to greater deal proneness, homeowners should be more deal prone than apartment dwellers.

The Effect of Income on Deal Proneness

The usual argument given in support of a negative relationship between deal proneness and income is that low income households have lower time costs resulting in lower search and transaction costs. Furthermore, economic theory suggests that lower income households should be more price sensitive. Empirical research in marketing has rarely shown that income affects deal proneness (see Webster, 1965, for example). If an effect is found at all, higher income seems to lead to greater deal proneness rather than less.

The problem with studying the effects of income is that income effects are confounded by the effects of household resource variables. The influence of these other variables must be removed before the effect of income can be clearly evaluated. It was stated above that car and home ownership should increase deal proneness. Higher income households are more likely to buy capital goods such as cars and homes, thus increasing their deal proneness. The resulting interaction between income and household resources may result in the anomalous finding that high income households are more deal prone than low income households. If resources available were held constant, however, one should observe the opposite outcome.


The data used to analyze deal proneness were the Chicago Tribune Panel purchase data and associated demographic variables. Consumers classified into three segments defined by Blattberg and Sen (1976): the National Brand Loyal Deal, the National Brand Switcher Deal, and Deal-Oriented, are defined here as being deal-oriented. All consumers classified into one of the other stable pat-term categories (i.e., not including the Changing Pattern or Last Purchase Loyal patterns) constituted the non-deal prone population. [The nonstable buying patterns contain some deal prone households who for certain periods of the data were deal prone and for other periods were not. They have been excluded because they were difficult to categorize. The size of this group is never more than 20% of the total consumers and is usually much smaller.] Five product categories were studied: aluminum foil, waxed paper, headache remedies, liquid detergent and facial tissue. The data were from 1958 to 1966, depending on the category. [Aluminum Foil (1962-66), Waxed Paper (1963-66), Liquid Detergent (1959-61), Facial Tissue (1958-61), and Headache Remedies (1959-61).] The household variables studied are those described in the previous section.

Blattberg and Sen (1976) classified each household's purchase patterns into segments. Deal proneness was based on membership in one of the above three segments and was a dichotomous variable: deal prone or not deal prone. Given that the effect of the independent variables on deal proneness may be nonlinear, it was decided to use cross-classification analysis instead of regression (see Frank, Massy and Wind, 1972, pp. 126-129). A major problem in doing cross-classification analysis is that sample sizes may become small for certain cells when two or three sets of independent variables are simultaneously analyzed along with the dependent variable. This is a particularly vexing problem here because most demographic and household resource variables are inter-correlated. For example, high income households who rented and did not own a car were a very small percentage of all high income households. Initially, the data were analyzed individually for each explanatory variable. Then certain combinations of variables were considered together. In future studies, if larger samples are available, interrelationships between more sets of explanatory variables should be studied.

Household Resource Variables

The first two variables analyzed were home ownership and car ownership. Tables 1 and 2 give the results for each of the five product categories. The table entries are the percentage deal prone. For example, in the case of waxed paper, 29.5% of the households who owned a home were deal prone and 12.5% of the households who did not own a home were deal prone.





Looking at the results in Tables 1-2, it appears that owning a car or a home makes a household much more deal prone. For every product category this result holds. These results are consistent with the predictions made in the previous section which indicated that home ownership and car ownership should lead to greater deal proneness.

One problem with analyzing the variables separately is that if one owns a car, one is also more likely to own a home. Thus, the observed effect may be due to one of the two variables and not the other. Table 3 studies the effect of the two variables jointly. The results show that except for facial tissue, owning both a home and a car results in the highest probability of being deal prone. The percentage deal prone is always higher when a household owned a car and a home than when it owned a car and rented. It is higher for 4 of the 5 products than when the household owned a home, but not a car. Thus, the effect is not due to only one of the two variables.



To get some idea of the magnitude of the effect that owning both a car and a home had on deal proneness, the following model was estimated:

Dij = abi gjeij     j=i,...,4  u=1,...,5     (1)


Dij = percentage of deal-oriented consumers for product category j and consumer characteristic i.

a = average deal-orientation.

bi = the effect of consumer characteristic i on deal-orientation.

gj = the effect of product category j on deal orientation.

eij = the disturbance term.

Because product categories with very low deal proneness may have a lower absolute difference in deal proneness for a given household characteristic than product categories with high deal proneness, a multiplicative model was used. If one takes logarithms of both sides of equation (1), the effects (bi and gj) can be measured using estimates calculated from standard analysis of variance formulas. The constraints are that

Ei ln(bi) = Ej ln(gj) = 0.

[This constraint is the same as requiring nibi = njgj = 1]

The model is similar to the log-linear models described by Green et. al. (1977) and Bishop et. al. (1975). The exact estimates used are given in Hogg and Craig (1970, pp. 327-334).

Estimates of the model's bi parameters are presented in Table 4. The results show that owning a car and a home had a 1.366 response, compared to not owning a car and renting, which had a .821 response. If we average across all the products so that the grand mean represents average deal responsiveness, we see that owning a car and a home increased deal responsiveness from 20.5% to 34.4%, a 67.9% increase. Owning either a car or a home, but not the other, increased deal responsiveness from 20.5% to 26.2%. It is clear, therefore, that owning both a car and a home greatly increases deal responsiveness, compared to not owning either or owning only a car or only a home.




On theoretical grounds, household income level should be negatively correlated with deal proneness. Empirically, the opposite relationship may be observed because of the effects of confounding variables. Table 5 gives the results of income for three income levels -- low ($0-5999) medium ($6000-8999) and upper ($9000 or more). The income categories are based on roughly 33% groupings. The results indicate that contrary to theoretical predictions, upper income households are more deal prone than low income households. For every product category except facial tissue, a higher percentage of upper income households are deal prone compared to low income households. The effect is not large in most categories, but it is persistent.



To isolate the effects of confounding variables, income was analyzed adjusting first for home ownership and then for car ownership. (Because cell sizes became very small, it was impossible to analyze income simultaneously with both home and car ownership). Tables 6-7 present the results. They indicate that if one adjusts for resources available, the income effect no longer persists (except for liquid detergent).





To see this more clearly, Tables 8-9 present the bi parameter estimates for the model in equation (1). First, for income and home ownership, the highest deal-oriented category is low income households that own a home. The difference between rents and owns for all three income groups is dramatic. The income effect seems to disappear when rent or own is taken into account. Similarly for car ownership, the income effect is small compared to whether a car is owned.





Thus, higher income does not lead to deal proneness as would be concluded if the income effect were not adjusted for car and home ownership. If the effects of car and home ownership could be simultaneously partialled out, we would expect to find even stronger evidence that lower income households are more deal prone than higher income households.


This paper has used a specific model of buyer behavior to identify household variables which should affect deal proneness. The model assumed that households make inventory decisions the same way that firms do. The variables that affect the buying decision are holding costs, stockout costs, transaction costs, purchase price and usage rates. Household characteristics were linked to these cost variables, and predictions were made about which types of households should be deal prone. The predictions were then tested empirically.

The empirical results showed that household resource variables: car and home ownership, were strong predictors of deal proneness. 34.4 percent of the households that owned a car and home were deal prone. Only 20.5 percent of the households that did not own either a car or a home were deal prone. The effects of income were also analyzed and the results showed that upper income households were more deal prone. However, when income was adjusted for household resources, this effect became negligible.


Becker, Gary S. "A Theory of the Allocation of Time," Economic Journal, 75 (September 1965), 493-517.

Bishop, Yvonne M. M., Stephen E. Fienberg and Paul W. Holland. Discrete Multivariate Analysis: Theory and Practice (Cambridge, Massachusetts: The M.I.T. Press, 1975).

Blattberg, Robert C., Thomas Buesing, Peter Peacock and Subrata K. Sen. "Identifying the Deal Prone Segment," Journal of Marketing Research (forthcoming, 1977).

Blattberg, Robert C. and Subrata K. Sen. "Market Segments and Stochastic Brand Choice Models," Journal of Marketing Research, 13 (February 1976), 34-45.

Frank, Ronald E., William F. Massy and Yoram Wind. Market Segmentation (Englewood-Cliffs, NJ: Prentice-Hall, 1972).

Green, Paul E., Frank J. Carmone and David P. Wachs-press. "On the Analysis of Qualitative Data in Marketing Research," Journal of Marketing Research, 14 (February 1977), 52-59.

Hillier, Frederick S. and Gerald J. Lieberman. Operations Research (San Francisco: Holden-Day, Inc., Second Edition, 1974).

Hogg, Robert V. and Allen T. Craig. Introduction to Mathematical Statistics (New York, NY: The Macmillan Company, Third Edition, 1970).

Kunreuther, Howard. "Why the Poor May Pay More for Food: Theoretical and Empirical Evidence," Journal of Business, 46 (July 1973), 368-83.

Michael, Robert T. and Gary S. Becker. "On the New Theory of Consumer Behavior," Swedish Journal of Economics, 75 (December 1973), 378-96.

Montgomery, David B. "Consumer Characteristics Associated with Dealing: An Empirical Example," Journal of Marketing Research, 8 (February 1971), 118-20.

Webster, Jr., Frederick E. "The 'Deal-Prone' Consumer,'' Journal of Marketing Research, 2 (May 1965), 186-89.



Robert Blattberg, University of Chicago (student), University of Chicago
Thomas Buesing, Wake Forest University
Peter Peacock, University of Rochester
Subrata Sen


NA - Advances in Consumer Research Volume 05 | 1978

Share Proceeding

Featured papers

See More


Turning the Titanic: Creating Consumer-Centric Cultures and Improved Consumer Experience in Large, Established Health Care Systems

Gregory Carpenter, Northwestern University, USA
Beth Leavenworth DuFault, University at Albany
Ashlee Humphreys, Northwestern University - Medill, USA
Lez Ecima Trujillo Torres, University of Illinois at Chicago, USA

Read More


C5. Krabby Patties, Kelp Chips, or KitKats?: Exploring the Depictions of Food Featured in Children’s Television Shows  

Kathy Tian, University of Illinois at Urbana-Champaign, USA
Regina Ahn, University of Illinois at Urbana-Champaign, USA
Michelle Renee Nelson, University of Illinois at Urbana-Champaign, USA

Read More


Machine Talk: How Conversational Chatbots Promote Brand Intimacy and Influence Consumer Choice

Thomas Hilden, University of Geneva, Switzerland
Christian Hildebrand, University of Geneva, Switzerland
Gerald Häubl, University of Alberta, Canada

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.