The Availability of Discretionary Time: Influences on Interactive Patterns of Consumer Shopping Behavior

ABSTRACT - This paper reports an empirical investigation of the relationships among consumer attributes (employment status, children status, education level, and income level) and shopping behavior which serves to reduce a consumer's inelastic time (increase discretionary time). An equilibrium theory of time expenditures forms the conceptual base for the paper. Log-linear multidimensional contingency table analysis is used to investigate the effects of interaction among consumer attributes and the effects of those attributes individually upon four shopping variables. As such, the analysis provides a methodological improvement over past research on the topic.

Citation:

Rebecca H. Holman and R. Dale Wilson (1980) ,"The Availability of Discretionary Time: Influences on Interactive Patterns of Consumer Shopping Behavior", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 431-436.

Advances in Consumer Research Volume 7, 1980     Pages 431-436

THE AVAILABILITY OF DISCRETIONARY TIME: INFLUENCES ON INTERACTIVE PATTERNS OF CONSUMER SHOPPING BEHAVIOR

Rebecca H. Holman, Pennsylvania State University

R. Dale Wilson, Pennsylvania State University

[The authors thank Stephen C. Rood, a former graduate student in marketing at Penn State, for his assistance with the preliminary phases of data analysis. Thanks are also in order to an anonymous reviewer for his/her especially helpful comments.]

ABSTRACT -

This paper reports an empirical investigation of the relationships among consumer attributes (employment status, children status, education level, and income level) and shopping behavior which serves to reduce a consumer's inelastic time (increase discretionary time). An equilibrium theory of time expenditures forms the conceptual base for the paper. Log-linear multidimensional contingency table analysis is used to investigate the effects of interaction among consumer attributes and the effects of those attributes individually upon four shopping variables. As such, the analysis provides a methodological improvement over past research on the topic.

INTRODUCTION

In recent years, scholars in a number of different disciplines have addressed the question of how consumers allocate time to shopping and other activities. Becker, an economist, was one of the first who explicitly incorporated time expenditures into representations of consumers' utility functions. In arguing for inclusion of time variables, Becker logically asserted the following:

For example, the cost of a service like the theatre or a good like meat is generally simply said to equal their market prices, yet everyone would agree that the theatre and even dining take time...time that often could have been used productively. If so, the full costs of these activities would equal the sum of market prices and the foregone value of the time used up (1965, p. 494).

Subsequent studies drawing upon Becker's theoretical positions have looked at such aspects of time allocation as the concept of leisure time and consumer preferences for leisure activities (Hawes 1977); differences in time expenditures by various groups of individuals (Szalai 1972); and the consumption of convenience products as a function of family life style variables (Anderson 1971). An excellent summary of these and other approaches to the study of time can be found in Jacoby, Szybillo, and Berning (1976).

One framework for understanding consumers' shopping behaviors in relation to time allocations stems from an equilibrium view of total time expenditures. If the ideal amount of discretionary time is not experienced, a consumer may strive to restore the equilibrium through a variety of shopping-related strategies. One purpose of the current paper is to examine conditions which are likely to result in consumers experiencing insufficient discretionary time (because of excessive fixed time), and then to determine if any of the postulated shopping-related strategies have been employed.

The other purpose of the paper involves the way in which inferences about the relationships among attributes of consumers and shopping behaviors are determined. Most of the past research on time expenditures and shopping behaviors has relied upon diary or survey data (Foote 1961). As most of these data are qualitative (categorical) in nature, past research has been restricted because the most sophisticated data analysis techniques require ordinal data at least. Recent innovations in methods for handling categorical data (e.g., Bishop, Fienberg, and Holland 1975; Goodman 1978) promise the ability to gain mere insight from these data than was possible in the past. The current paper uses this new method (log-linear contingency table analysis) to determine the effects of the interaction among independent variables, thereby illustrating the usefulness of such an approach to data manipulations.

CONCEPTUAL OVERVIEW

The literature dealing with time allocations by consumers stems mainly from Becker's seminal work, which was itself an outgrowth of economic studies of labor supply (Becker 1965, p. 494). Time was variously divided into work time and free time; time available for work and time available for leisure; and fixed time and discretionary time. As pointed out most recently by Hendrix, Kinnear, and Taylor (1979), but also of concern to others before them (Becker 1965, p. 504; Robinson 1977, p. 137-138; Phlips 1978, p. 1034), there are significant conceptual and operational difficulties in trying to devise a taxonomy of time allocations. The Hendrix, Kinnear, and Taylor operationalization hinges on the degree to which consumers' time is constrained externally. While they operationally distinguish three categories (inelastic, intermediate, discretionary), they state that the three more precisely form a continuum ranging from time totally controlled externally (work time on an assembly line for example) to time totally under internal control (the traditional concept of free or leisure time). It is the Hendrix, Kinnear, and Taylor conceptualization that is used for the balance of this paper.

A second concept, whose genesis was with Becker and the economists who followed him, concerns the substitutability of time, goods, and money. Becker argued the substitutability of the three (1965, p. 498), but Phlips explicitly rejected Becker's position stating that "...there are not special links between leisure and money in the preference ordering" (1978, p. 1026). Further arguments criticizing the assumption of total latitude in time, goods, and money transfers are summarized by Hendrix, Kinnear, and Taylor (1979). While such substitutions may be found through empirical investigations (as did Phlips for time and work) it is not axiomatic that such choices will be made by consumers. [Another interesting substitution problem involving time is discussed by Mabry (1970): time and money are again two of the variables, but stamina is the third in this conceptualization.]

That consumers demand goods and services that lead to an increase in discretionary time has been investigated empirically, and confirmed by Marple and Wissman (1968) and Anderson (1971). A theoretical statement regarding how the marketplace can provide time utility can be found in Kelley (1958), who quoted Mortimer's ten convenience forms (Mortimer 1955, p. 7-17). More recently, Coyle, in discussing how retail grocers could (and should) react to the success of fast-food restaurants stated that convenience "is where shoppers would like to see the most improvement'' (1977, p. 38). Therefore, it appears that Becker's conceptualization was not incorrect for some situations, and that goods and money can be used by consumers to acquire greater amounts of discretionary time.

Another major body of work, which was predicted by Becker (1965, p. 517), is the collection of time budget data from consumers and the subsequent identification of differences in time usage by different groups of people. The largest time budget study was reported by Szalai (1972). This and other similar research projects are thoroughly reviewed in Venkatesan and Arndt (forthcoming).

The most fruitful outcome of time budgets is the realization that consumers allocate time to various activities as a function of a number of factors. The theoretical basis for research which studies differences in time allocations is the recognition that (1) some consumers may experience less discretionary time than others and (2) these consumers may wish to increase their total discretionary time by decreasing inelastic or intermediate time. If it can be assumed that individuals desire to maintain some sort of optimal relationship between discretionary time and all other time, an assumption which is implicit in Phlips' work (1978, p. 1027), then it follows that those individuals who experience less discretionary time than they desire will engage in efforts to restore a balance that is closer to their individual ideal states.

The purchase of goods and services offering convenience constitutes one strategy for increasing discretionary time, if it is assumed that shopping activities are seen as either inelastic or intermediate time expenditures, an assumption with some support in the literature. The international time budget study so-conceptualized shopping time categorizing it as part of "errands", a subset of "other household obligations" (Szalai 1972, p. 565). Additionally, Hendrix (1979) found that "grocery shopping" was ranked 21 and "other kinds of shopping" was ranked 18 in a list of 22 activities, where 22 was the least desirable. Clearly, though, some types of shopping can be seen as discretionary by particular groups of consumers, so that the assumption that shopping is an onerous task is undoubtedly an oversimplification.

Empirical evidence indicates that the perception of shopping as something other than a discretionary time expenditure may be a function of certain family structural and other variables. (Representative research includes Reid 1963; Anderson 1971, 1972; Prochaska and Schrimper 1973; Douglas 1976; Arndt and Gr°nmo 1977; and McCall 1977.) Results from this research have been mixed and often contradictory. For example, Anderson (1972) did not find a significant relationship between employment of women and convenience food orientation, but did find statistical significance for stage in family life cycle, family size, age of household head, and education of household head. By contrast, Arndt and Gr°nmo (1977) found a positive relationship among total time spent shopping, the employment status of the woman, and stage in the family life cycle, but no relationship between shopping time and number of children in the family.

The theoretical basis for studying family structural variables and hypothesizing that some of these would tend to place pressures upon individuals to decrease shopping and other non-discretionary time expenditures in order to reestablish optimal levels of discretionary time is also supported by Stampfl (1978). Stampfl, who characterized a variety of "consumer elements" through thirteen stages of the life cycle, sees time as a different constraint at different stages (1978, p. 215). Stampfl's work, as well as that of the other researchers cited above, suggest a fruitful way of conceptualizing individuals most likely to need to increase discretionary time due to excessive time commitments. Women who work outside the home or have children are likely to experience the smallest proportion of discretionary time, and would be expected to use more strategies for decreasing the total time spent on shopping, than would women who do not work outside the home or have children. Furthermore, higher income and higher education are likely to be enabling factors in consumers' decisions to engage in shopping behaviors designed to reduce shopping time: those with greater income can afford to pay for convenience; those with greater years of education may be better able to evaluate alternative shopping strategies to arrive at the minimum possible total shopping time.

It is important to consider the interactions among these variables as it is the combinations among them that are likely to be significant at the extremes. Therefore a woman who works outside the home and has children and has higher income and has higher education is the most likely user of strategies for reducing shopping to an optimal level both because she feels the pressure to do so, and because she is able to do so. By contrast, a consumer who does not work outside the home, has no children, has lower income, and has lower education is likely to be the consumer not using strategies for reducing total shopping time, both because she need not do so, and because she is less able to do so.

The research presented here specifically looks at the interaction among these variables and as such is an improvement over past research since these interactions may point to relationships not uncovered in past research. In fact, only two of the previous studies of time usage have employed a methodology that could detect interactions. Anderson (1971) implicitly [Anderson first clustered subjects on the basis of convenience goods orientation. Then chi-square tests determined significant differences between clusters.] considered such interactions, while Douglas (1976) explicitly dealt with them. All others ignored the effects of the interactions among independent variables upon dependent variables. It was with this shortcoming in mind that the research question for the present study was formulated:

Are family attributes (employment status of female head of household, presence of children, income, education) individually and/or jointly related to incidence of shopping behaviors which reduce total time spent on shopping?

If this research question is answered affirmatively, then the research presented here has potential for wide significance because it provides not only some concrete results, but also a procedure for pursuing further investigations of these relationships using diary and survey data.

METHODOLOGY

Overview

In an attempt to provide an answer to the research question, a methodology was designed that would evaluate the effects (both main effects and interaction effects) of four family structural variables on each of four measures of grocery store shopping activities. These consumer shopping activities, which are the endogenous variables used in the study, all reflect strategies for reducing the amount of time spent in shopping-related activities. The methodology uses multidimensional contingency table analysis to test the impact of employment, the presence of children, education, and income on grocery shopping behavior patterns that are likely to result in time saving to the female head of household.

The Data

In order to provide an empirical test of the effects of family structural variables on consumer shopping behavior, a data tape reporting the results of a study on grocery shopping behavior was obtained. This data set, consisting of self-reported shopping behavior during a one-month period in the fall of 1970, contained consumer diary panel records for 719 household in the Chicago, Illinois, standard metropolitan statistical area (SMSA). The panel records report total grocery store expenditures in 14 food categories and four non-food categories, as well as the specific store chosen, the specific time period in which the shopping was conducted, and the specific individual household and non-household member(s) who made the shopping trip. In addition, a large number of demographic and socio-economic characteristics were available for each participating household. This panel study was designed to accurately represent the behavior of the U.S. urban population via probability sampling methods. [As an anonymous reviewer pointed out, the panel composition probably was not strictly "urban" by census definition and could probably be better described as representing the adult, married, middle-class, noninstitutionalized, continental, non-military urban population.] The data were collected by a well-known supplier of commercial market research data and the collection of the data was co-sponsored by a large regional chain of supermarkets and a large national producer and distributor of consumer food products.

Of the 719 panel respondents comprising the grocery shopping study, 35 were unmarried and were dropped from the analysis since their proportion of the total sample was extremely small (4.87%) and since the bulk of previous research has dealt exclusively with married females. Two respondents were also dropped from the analysis because they failed to report their level of education. Thus, 682 households comprise the final sample.

Operationalization of Behavioral Measures

Four dependent variables were computed directly from the shopping behavior data. These variables, which represent different aspects of family shopping behavior, consist of the percentage for each household relative to the total number of: (1) shopping trips made to smaller, convenience-type grocery stores [These convenience-type stores were identified in the codebook prepared by the commercial data supplier.] (as opposed to larger supermarkets offering a wider range of products); (2) "filler" shopping trips (as opposed to more "major" shopping trips); (3) trips made in seven "day/time" periods in which the panel reported making few shopping trips; and (4) trips made by someone other than the female head-of-household. These dependent variables were designed as surrogates for "total time spent shopping" and, as such, probably do not correlate one-to-one with total time spent shopping.

For the purposes of this research, "filler" shopping trips were defined as those trips in which the household spent $5.00 or less on both food and non-food items. While this $5.00 cut-off is somewhat arbitrary, Wilson (1977, p. 62-64) provides evidence that it is appropriate for the shopping behavior data used here. The seven "day/time" periods in which few shopping trips were made were identified from the joint frequency distribution of the total trips made over the duration of the data collection period. Of the 21 possible "day/time" periods (i.e., seven days of the week and three time periods--morning, afternoon, and evening), the seven "day/time" periods studied represented only 18.4% of the total number of shopping trips. [These seven low-volume "day/time" periods were Sunday evening, Saturday evening, Sunday morning, Sunday afternoon, Tuesday evening, Wednesday evening, and Monday evening. Interestingly, all of these periods occur on weekends or during weekday evenings and, therefore, would be accessible to female panel respondents who are employed during weekday mornings and afternoons.] As such, the consumer is much less likely to encounter average- or high-volume store traffic during these seven low-volume periods in contrast with the remaining 14 "day/ time" periods.

Of the four dependent variables studied in this paper, two variables (percentage of filler trips and percentage of trips made in low-volume store traffic periods) are unique to this research. The two other dependent variables (percentage of trips to convenience stores and percentage of trips made by someone other than the female head-of-household) were calculated to correspond to the dependent variables used by other researchers who have studied consumers' time-orientation (e.g., Anderson 1971; Douglas 1976).

The independent variables used to explain the dependent variables were also taken directly from the shopping study. Here, four family structural variables were taken from the demographic/socio-economic report for each household. These variables include the employment status of the female respondent (employed vs. unemployed), whether children are present in the household (children vs. no children), educational status of the female respondent (did not attend college vs. attended college), and income status of the household (low income vs. high income). These variables are representative of those used by other researchers.

Table 1 was prepared to further define and summarize the operationalization of the independent variables and to indicate the differences between the dichotomous groups as defined by the independent variables across the four dependent variables. For the most part, Table 1 indicates that the means are in the expected direction; [The past literature was used to establish a priori expectations concerning the directionality of the means for each combination of independent and dependent variables. Those cases that violate these expectations are noted in Table 1. In the case of income status and percentage of trips made during low volume store traffic periods, no directionality could be predetermined and a two-tailed t-test was used. In all other cases, a one-tailed test was used.] and the differences between means in five of the 16 independent variables combinations (31.25%) are significant at the .05 level or less. However, since the interactions among independent variables are not considered in the data presented in Table 1, these data provide little insight into the full relationship between the set of independent variables and each dependent variable. Similarly, a large proportion of the data analysis methods that have been used in past research in attempting to assess the effects of time pressures on consumer behavior have also failed to examine potentially interesting interactions among variables.

Method of Analysis

In an attempt to provide a meaningful understanding of the impact of the independent variables on consumers' shopping behavior and thus contribute to consumer behavior theory, multidimensional contingency table analysis was chosen as the method of analysis. The use of multidimensional contingency table (log-linear) analysis is appropriate since the independent variables are categorical and the concern of the research is whether these variables are individually and/or jointly associated with each of the dependent variables (see Davis 1974; Bishop, Fienberg, and Holland 1975, and Goodman 1978). The advantage of the general log-linear model over other methods is that it can conveniently test all main effects and all possible first- and high-order interactions by determining the degree of goodness-of-fit associated with these main effects and interactions. As such, these procedures are interpreted similarly to analysis of variance, except that the log-linear model is designed for use with cell frequencies or cell proportions. The use of the log-linear model is especially advantageous in analyzing the data in the current study because the collection of the data was not experimentally controlled, and, therefore, cell sizes vary widely. The data were analyzed by a general linear model data analysis package (Scott 1975) which includes a subroutine for analyzing multidimensional contingency tables.

TABLE 1

OPERATIONAL MEASURES AND SUMMARY DATA

Because of the interest in assessing the effects of the interactions among the independent variables, a "saturated model" was developed for each dependent variable. The saturated model includes all maim effects and all possible interactions and, as a consequence, has as many parameters as the contingency table has cells (Bishop, Fienberg, and Holland 1975, p. 9). Other models (termed "adjacent models") were then developed, each of which removed one successive term from the saturated model. Each of these successive nested models was evaluated relative to a more complex adjacent model (which had one additional term) by considering the incremental informational content contained in the simpler adjacent model. This procedure, advocated by Ku and Kullback (1968; 1974), uses the likelihood ratio statistic, G², as a measure of goodness-of-fit (see Bishop, Fienberg, and Holland 1975, p. 124-131). The likelihood ratio statistic is defined similarly to the common Pearson c² and, in fact, is distributed asymptotically as c² with n₁ degrees of freedom. From all indications, the incremental informational content approach to model evaluation appears to he the most attractive method for the data analyzed here, especially in light of Everitt's (1977, p. 84) implicit suggestion that other methods of model evaluation (see Bishop, Fienberg, and Holland 1975, p. 165-168) may involve improper procedures.

In the case considered here, the saturated model consists of a "four factor, one response model," where the four factors refer to the four independent variables (each of which has two levels) and the one response refers to the classification of the dependent variable under consideration in relation to the independent variables. As indicated previously in Table 1, independent variables A, B, C, and D are defined dichotomously. The input values for the dependent variables (considered separately in this analysis) are defined as the mean percentages for each cell in the multidimensional contingency table.

Operational Hypotheses

In line with the conceptual review and the discussion of the saturated log-linear model presented earlier, it is expected that the effects of the independent variables would be reflected in actual shopping behavior. In particular, the null hypotheses to be tested are that each successive model term contributes nothing to the explanation of the dependent variable under consideration. Likewise, the alternative hypotheses state that each independent variable, considered individually and in conjunction with other independent variables, does contribute to the explanation of the dependent variable. These hypotheses are tested by evaluating the magnitude of the incremental value of the likelihood ratio statistic as each successive term of the saturated model is eliminated. Since the likelihood ratio statistic can conveniently be partitioned among nested models (see Bishop, Fienberg, and Holland 1975, p. 125-130), its incremental difference is distributed as c² with n₂ - n_l degrees of freedom (where n₂ is the degrees of freedom for the simple model and n₁ is the degrees of freedom for the more complex model). The incremental-difference likelihood ratio statistic can then be compared to a conventional chi-square table to determine the appropriate level of statistical significance.

RESULTS AND DISCUSSION

The results of the study are summarized in Table 2, which displays the information contribution for each of the 15 terms in the saturated model. As the table indicates, the information contribution is statistically significant at the .05 level or less for all main effects and interaction effects, and this finding is consistent across all four dependent variables. In 57 of the 60 instances (95.00%) in which the incremental G² statistic was evaluated, the information contribution is significant at the .001 level. These results indicate that each independent variable, when considered individually and when considered jointly with one or more of the other independent variables, provides a meaningful contribution to the dependent variables.

Of particular importance in determining the overall effects of employment status, presence of one or more children, educational status, and income status on the store shopping behaviors used as dependent variables is the relationship of the "Explained" G² to the "Error" G². The "Explained" G² is the total information contributed by all main and interaction effects, and it can be determined by summing the G² statistics for each model component. The "Error" G² was determined by testing the equi-probable model (that is, all cells have the same underlying theoretical probability). When the "Explained" G² is divided by the "Error" G², the resulting percentage can be thought of as the percentage of total variation in the multidimensional contingency table that is explained by the independent variables.

As Table 2 indicates, this percentage of variation for the saturated model that is explained for (1) the percentage of trips made during low volume hours and (2) the percentage of trips made by someone other than the female head-of-household is low (11.71% and 26.65%, respectively). However, for the remaining two dependent variables--percentage of trips to convenience stores and percentage of filler trips--the independent variables in the saturated model explain extremely large percentages of the variation in the data (95.03% and 95.84%, respectively). It should be noted that although the analysis presented in this paper reveals interesting relationships between family structural variables and consumer strategies for reducing time expenditures for grocery store shopping, the analysis is somewhat limited by a lack of generalizability. This is due to small sample sizes in two cells (employed/no children/attended college/low income and not employed/no children/attended college/low income) in the multidimensional contingency table. These cell sizes are not small enough to cause statistical or interpretation problems, but they do limit the conclusions that can be drawn from the data. [Because of space limitations, the Results and Discussion section is abbreviated. The complete version of the paper is available from the authors.]

TABLE 2

INFORMATION CONTRIBUTION (G²) FOR EACH MODEL TERM

REFERENCES

Anderson, W. Thomas, Jr. (1971), "Identifying the Convenience-Oriented Consumer," Journal of Marketing Research, 8, 179-183.

Anderson, W. Thomas, Jr. (1972), "Convenience Orientation and Consumer Behavior,'' Journal of Retailing, 48, 49-71.

Arndt, Johan and Gronmo, Sigmund (1977), "The Time Dimension of Shopping Behavior: Some Empirical Findings," in Advances in Consumer Research, Vol. IV, ed. William D. Perreault, Jr., Atlanta: Association for Consumer Research.

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

Bishop, M. M., Fienberg, Stephen E., and Holland, Paul W. (1975), Discrete Multivariate Analysis: Theory and Practice, Cambridge, MA: The MIT Press.

Coyle, Joseph S. (1977), "The Battle for Eating In," Progressive Grocer, 35-41.

Davis, James A. (1974), "Hierarchical Models for Significance Tests in Multivariate Contingency Tables: An Exegesis of Goodman's Recent Papers," in Sociological Methodology 1973-74, ed. Herbert L. Costner, San Francisco: Jossey-Bass.

Douglas, Susan P. (1976), "Cross-National Comparisons and Consumer Stereotypes: A Case Study of Working and Non-Working Women in the U.S. and France," Journal of Consumer Research, 3, 12-30.

Everitt, B. S. (1977), The Analysis of Contingency Tables, New York: Halsted Press.

Foote, Nelson N. (1961), "Methods for Study of Meaning in Use of Time," in Aging and Leisure: A Research Perspective into the Meaningful Use of Time, ed. Robert W. Kleemeier, New York: Oxford University Press.

Goodman, Leo A. (1978), Analyzing Qualitative/Categorical Data, Cambridge, MA: Abt Books.

Hawes, Douglass K. (1977), "Time Budgets and Consumer Leisure-Time Behavior," in Advances in Consumer Research, Vol. IV, ed. William D. Perreault, Jr. Atlanta: Association for Consumer Research.

Hendrix, Philip E. (1979), "The Role of Positive Reinforcement in the Consumer's Allocation of Time," unpublished manuscript, Florida State University.

Hendrix, Philip E., Kinnear, Thomas C., and Taylor, James R. (1979), "The Allocation of Time by Consumers," in Advances in Consumer Research, Vol. VI, ed. William L. Wilkie, Miami: Association for Consumer Research.

Jacoby, Jacob, Szybillo, George J., and Berning, Carol Kohn (1976), "Time and Consumer Behavior: An Interdisciplinary Overview," Journal of Consumer Research, 2, 320-339.

Kelley, Eugene J. (1958), "The Importance of Convenience in Consumer Purchasing," Journal of Marketing, 23, 32-38.

Ku, Harry H. and Kullback, Solomon (1968), "Interaction in Multidimensional Contingency Tables: An Information Theoretic Approach," Journal of Research of the National Bureau of Standards, 72B, 159-199.

Ku, Harry H. and Kullback, Solomon (1974), "Logit Models in Contingency Table Analysis," The American Statistician, 28, 115-122.

Mabry, Bevars D. (1970), "An Analysis of Work and Other Constraints on Choices of Activities," Western Economic Journal, 8, 213-225.

Marple, Gary A. and Wissmann, Harry B., eds. (1968), Grocery Manufacturing in the United States, New York: Frederick A. Praeger, Publishers.

McCall, Suzanne (1977), "Meet the Workwife," Journal of Marketing, 41, 55-65.

Mortimer, Charles J. (1965), Two Keys to Modern Marketing, Scarsdale, NY: The Updegraff Press, Ltd.

Phlips, Louis (1978), "The Demand for Leisure and Money," Econometrica, 46, 1025-1043.

Prochaska, Fred J. and Schrimper, R. A. (1973), "Opportunity Cost of Time and Other Socioeconomic Effects on Away-From-Home Food Consumption," American Journal of Agricultural Economics, 55, 595-603.

Reid, Margaret (1963), "Consumer Response to the Relative Price of Store Versus Delivered Milk," Journal of Political Economy, 71, 180-186.

Robinson, John P. (1977), How Americans Use Time: A So-cial-Psychological Analysis of Everyday Behavior, New York: Praeger.

Scott, Del T. (1975), "RUMMAGE: A General Linear Models Data Analysis System," unpublished manuscript, Computation Center Contribution Program, The Pennsylvania State University.

Stampfl, Ronald W. (1978), "The Consumer Life Cycle," The Journal of Consumer Affairs, 12, 209-219.

Szalai, Alexander, ed. (1972), The Use of Time, The Hague: Mouton.

Venkatesan, M. and Arndt, Johan (forthcoming), "Temporal Dimensions of Consumer Behavior: An Explanation With Time-Budget," in Research in Marketing, ed. Jagdish N. Sheth, Greenwich, CT: JAI Press.

Wilson, R. Dale (1977), "Generalized and Embedded Versions of Heterogeneous Stochastic Models of Consumer Choice Behavior: An Empirical Test and Statistical Evaluation in a Dynamic Store Selection Context," unpublished doctoral dissertation, Department of Business Administration, The University of Iowa.

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