Time Allocation Dimensions of Shopping Behavior

R. Dale Wilson, Cornell University
Rebecca H. Holman, Young & Rubicam, Inc.
ABSTRACT - Much existing time allocation research uses an economic approach. This article extends the traditional economic concept of time by utilizing a psychologically-based theory of time equilibrium. A descriptive model of time allocation which incorporates this theory is tested in the context of consumer shopping behavior.
[ to cite ]:
R. Dale Wilson and Rebecca H. Holman (1984) ,"Time Allocation Dimensions of Shopping Behavior", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 29-34.

Advances in Consumer Research Volume 11, 1984      Pages 29-34

TIME ALLOCATION DIMENSIONS OF SHOPPING BEHAVIOR

R. Dale Wilson, Cornell University

Rebecca H. Holman, Young & Rubicam, Inc.

ABSTRACT -

Much existing time allocation research uses an economic approach. This article extends the traditional economic concept of time by utilizing a psychologically-based theory of time equilibrium. A descriptive model of time allocation which incorporates this theory is tested in the context of consumer shopping behavior.

INTRODUCTION

Time, viewed as an economic exchange medium, has attracted the attention of scholars in a wide variety of disciplines as reflected by the significant number of works that assume this stance (e.g., Becker 1965; Abbott and Ashenfelter 1976; Vickery 1977; Phlips 1978; Voss and Blackwell 1979). The implication is that time, like money and other commodities, can be spent, traded, borrowed, bought, sold, and saved. Unlike money, time is a fixed-budget commodity; one is never able to accumulate any more than a certain quantity of time during a particular period and it must be spent, even if one makes no overt effort to do so. There are no true time "savings" as there are money savings because one cannot access time that was "saved" in the past. One's allotment of time for any given period is always fully consumed at the end of that period.

The common phrase "time savings" refers more precisely to a decrease in time expenditures for one event so that some other (presumably more desired) event may occur. Study of the phenomenon of shifting time allocation among activities has led economists (e.g., Becker 1965) to incorporate time expenditures into consumers' utility functions. A logical out-growth of Becker's work is the study of how different individuals budget their time (Szalai 1972; Robinson 1977).

Any study of time allocation must cope with two key issues: the categories of events from which individuals attempt to "borrow" time and the corresponding categories of activities to which that time is subsequently allocated; and the reasons for such shifts of time expenditures. The substitutability of time, goods, and money has been proposed as one way of dealing with these issues (e.g., Becker 1965; Vickery 1977). When the marginal utility of money exceeds the marginal utility of goods and/or time, one engages in some sort of compensatory behavior, namely foregone goods or incremental work hours. Work time and non-work (i.e., leisure) time are seen as complementary, forming the time-allocation categories used by economists.

The economic position has been criticized on several bases. The assumption of the substitutability of time, goods, and money has been challenged (e.g., Phlips 1978), although there is limited empirical evidence that such substitutions may actually occur (Prochaska and Schrimper 1973; Abbott and Ashenfelter 1976). Of more concern is the assertion that the time-allocation categories of economics are not relevant for understanding today's consumer. Voss and Blackwell (1979) point out that individuals can augment their time budgets in more ways than simply by reducing work time. However, workers no longer have the degree of discretion over their work hours that they once did. Other categories of time may be more germaine to the consumer than "work" and "leisure".

A third criticism of the economic position is the most important from the perspective of the consumer behavior theorist. The economic explanation for why and under what circumstances time-allocation substitutions are made lacks the kind of theoretical sophistication that would lead to hypothesis generation and testing regarding differences in consumption patterns (cf. Hendrix, Kinnear, and Taylor 1979). Due to the oft-stated difficulties in operationally defining "utility", it is doubtful that the economic model can be of much use to the researcher wishing to explain differences in time-allocation behavior.

This article attempts to overcome the problems associated with the economic approach by empirically testing a descriptive model of time allocation that was developed by Holman and Wilson (1982). A taxonomy of time allocation developed by Chapin (1974) forms the foundation for the model, although Chapin's approach is expanded by introduction of an equilibrium theory of time expenditures based on psychological rather than economic need satisfaction. The model used here has the advantage of predicting specific time allocation responses. Four such responses, relating to grocery shopping, are examined in providing an empirical test of a portion of the model

CONCEPTUAL OVERVIEW

Chapin (1974) envisaged arranging behaviors along "a continuum extending from obligatory to discretionary forms of activity" (p. 37). Activities at the obligatory end are those for which the individual is constrained by some physiological or social need (e.g., sleeping, child care). This is not to say that the individual has no choice about the time to allocate to these activities, especially in the short run, but rather that there is less freedom to choose than for activities at the other end of the continuum. Discretionary activities are those for which one has more personal control over the time allocated than the portion of one's time budget dictated by social or physiological pressures (e.g., recreational events, hobbies). While Chapin would classify some activities as purely obligatory or purely discretionary, the remainder would fall somewhere in between and their placement along the continuum would "vary with the person and the circumstances" (Chapin 1974, p. 37). It is important to note that the exact location of a particular activity cannot be known a priori, but must be determined empirically.

Implicit in much of the previous work on time allocation is the theory that individuals desire to maintain an optimal mix of obligatory and discretionary activities (cf. Holman and Wilson 1982). Such a desire is distinct from the economic concept of opportunity costs for foregone activities, stemming rather from a number of needs (e.g., need for structure, need for variety, the Protestant work ethic). When individuals are faced with an imbalance of either type of activity (when the mix of obligatory and discretionary activities exceeds some tolerable range about an ideal point), there will be some compensatory activitY to restore equilibrium.

Much empirical research indicates that the character of one's household is a determinant of one's time expenditures, especially as the presence of others could result in social obligations that decrease the time available for discretionary activities. Employment status of the female head of household is one of these household characteristics that has received considerable attention, especially in studies focussing on the female as the buyer (e.g., Prochaska and Schrimper 1973; Strober and Weinberg 1977). Justification for attention to employment status of the female head derives from the historical interest in work time and the changing roles of women in addition to the fact that employment is often seen as an obligatory activity close to the end of the time allocation continuum (supplanting discretionary as well as obligatory activities, Vickery 1977). Family life cycle is another important household characteristic (Stampfl 1978) as it accounts for the presence of others in the household and suggests the demands those others might place upon the time of household heads (especially the female).

There are certain household characteristics that enable people to engage in compensatory behavior should a disequilibrium occur. For example, some of the compensatory activities that decrease obligatory time are quite expensive (e.g., in-home child care); others require better education (e.g., higher-paying jobs because of better skills thereby allowing a decrease in work hours for equal compensation). Both education and income (wealth) are important facilitators.

The background material presented above enables the description of an equilibrium model of time allocation as illustrated in Figure A (and discussed in detail by Holman and Wilson 1982). Household characteristics form the conditions which lead to disequilibrium in time allocation although some of these characteristics enter the model as enabling factors to equilibrium restoration rather than causing disequilibrium.

FIGURE A

AN EQUILIBRIUM MODEL OF TIME ALLOCATION AND SHOPPING BEHAVIOR

AN EMPIRICAL TEST OF THE MODEL

The descriptive model of time allocation was tested by examining four grocery shopping behaviors. One of these (purchase of convenience goods) is a precursor activity to a non-shopping strategy (cooking) for reducing the total time allocated to food preparation. The other three are shopping strategies that in and of themselves reduce total shopping time. First, the purchase of non-food items (e.g., small tools, non-prescription drugs) in a grocery store eliminates the necessity of making a special trip to stores specializing in those items. Second, through the purchase of a small quantity of items that satisfy immediate needs, one is able to defer more time-consuming major trips until a point in time characterized by fewer pressures to reduce obligatory activities. Third, the practice of sending others (e.g., children, neighbors) to the grocery store allows the female head of household to effectively avoid time expenditures in grocery shopping.

An earlier study by Holman and Wilson (1982) used two other types of behaviors to test the model used here. That application indicated that consumers acted in accordance with the model by shopping in grocery stores during non-busy hours in order to save time. However, a second dependent variable, the choice of convenience stores instead of full-line supermarkets, was not satisfactorily explained by the model. Using the model with four-additional dependent variables, each of which was defined to investigate separate time-saving strategies, provides a more thorough test of the conceptual model as well as allowing additional insights into the process by which individuals balance obligatory and discretionary time expenditures.

In order to test the nature of the relationship between the household characteristics suggested by the model and four grocery shopping behaviors, a two-phase data analysis methodology was adopted. This methodology was expected to provide important insights into the question of whether the equilibrium model of time allocation yields an accurate description of time consumption in shopping behavior.

Methodology

A methodology which evaluates both main and interaction effects of the household characteristic variables on consumer shopping patterns was used to test the hypotheses stemming from the equilibrium model of time allocation. The methodology, which employs a combination of Automatic Interaction Detector (AID) and Multiple Classification Analysis (MCA) is the same as that used by Holman and Wilson (1982).

The data used in the study contained self-recorded grocery store expenditures collected in the fall of 1970 for fourteen food and four non-food categories, information on the specific store where the purchases were made, the specific household and non-household member(s) who made the trip, and a large number of demographic and socio-economic characteristics for the participating households. A total sample of 709 households (each of which was located in the Chicago metropolitan area) was used for subsequent analysis; twelve households were dropped because of incomplete data in their demographic records.

Four dependent variables were calculated directly from the shopping behavior data. Two of these variables were computed as proportions of the total number of trips made by each household during a four-week period for the following types of consumer shopping strategies: the number of "filler" shopping trips and the number of shopping trips made by family or non-family member(s) other than the female head of household (referred to as "trips by others"). The remaining two dependent variables were computed as proportions of the total dollar expenditures on foot and non-food items made by each household during the four-week period. These included the total dollar expenditures on non-food items and the total dollar expenditures in five convenience food categories. Because of the wide variation both in the number of trips per household (X = 14.97; a = 6.85) and the total dollar expenditures per household (X = 5132.94; a = $55.66), all four dependent variables were computed as proportions. This procedure serves to create a common ground upon which the shopping behaviors of each household may be evaluated. Otherwise, it would be impossible to isolate the effects of each household's shopping strategy from measurement artifacts caused by variations in their trip frequencies and expenditure levels.

For the data used in this study, "filler trips" are defined as those trips in which the household spent five dollars or less on both food and non-food items. While the five dollar cut-off is somewhat arbitrary, Wilson (1977, p. 62-4) provides evidence that it is appropriate for the data used her An analysis indicated that as the level of expenditures nears the five-dollar figure, the shopping strategy used by many households seems to change from purchasing a limited assortment of regularly purchased and consumer staple items (e.g., bread, milk) to a wider assortment of times that are consumed and purchased less frequently (e.g., varieties of fresh meats, fresh fruits and vegetables). The feature that distinguishes "filler" and "major" trips seems to be the overall depth and variety of items purchased. This finding is in line with MacKay's (1973) analysis which suggests the likelihood of differences between consumers' mental choice processes for major and minor trips. The shopping strategy for minor trips may be dominated by convenience considerations while other factors are likely to exert influence on the decision process for major trips.

The only other dependent variable requiring further explanation is convenience food purchases. The five food categories comprising these expenditures include convenience-oriented frozen foods other than frozen fruits and vegetables (e.g., frozen juices, baked goods); canned or packaged convenience foods (other than canned fruits, vegetables, and meats); an three categories of ready-to-eat and delicatessen items (e.g., salads, pre-cooked meats, other deli items). Because the specific contents of each food category were pre-coded at the time of the original collection of the data, it was not possible to redefine the categories to eliminate those items that may be universally consumed and therefore not necessarily associated with a time saving strategy. As a result there is reason to believe that "convenience food" purchase as defined here may be confounded with foods that are normally thought of as less convenience oriented.

The independent variables that were used to explain the dependent variables represent the subset of demographic and socio-economic characteristics that was expected to have so potential impact upon time allocation as indicated in Figure A. Although hypotheses have been developed concerning the expected relationships between each dependent variable and small group of independent variables, the larger group of variables (eleven in all) was used in the first phase of the data analysis (in which AID was employed). The AID procedure was used for the express purpose of identifying any interaction effects occurring among the independent variables of interest and other related demographic and/or socio-economic characteristics. In this phase of the analysis, the following set of discretely-categorized variables was studied: t age, marital status, employment status, and educational status (in number of years) of the female head of household; the employment status of the male head of household; the type of dwelling unit occupied by the household; the current market value of the home (if owned); the total household size; the age distribution of members of the household other than the male and female heads; the number of automobiles consistent available to the household; and the household's total income. If interaction effects were found, the appropriate interaction term(s) was then included in the second phase of the study. A combination of the AID3 version of the Automatic Interaction Detector (Sonquist, Baker, and Morgan 1973) and MCA (Andrews et al. 1973) combine to provide a powerful the of the relationship between these household characteristics and shopping-related activities that may yield time savings. A more complete discussion of these methods is available in Holman and Wilson (1982).

Table 1 presents variable definitions of the seven independent variables for which hypotheses were generated. In accordance with these hypotheses, each independent variable" entered into the MCA model for the appropriate dependent variable(s). Based on the outcome of the AID3 analysis, the criterion variables were entered into each MCA model as either a main or in interaction effect.

TABLE 1

VARIABLE DEFINITIONS

Because each dependent variable was stated as a proportion of the total number of shopping trips or as a proportion of the total grocery store expenditures, the dependent variables were transformed via an arc sine transformation (see Winer 1971, p. 399-400; Neter and Wasserman 1974, p. 507-8). The MCA models were also run with the non-transformed proportions so that easily-interpretable class means could be reported. No important differences were found among the transformed and non-transformed sets of results.

RESULTS AND DISCUSSION

The AID3 runs confirmed the expectation that significant interactions existed in the data. Interactions were detected for each dependent variable except for non-food purchases. These interactionS are detailed in Table 2.

The results of the AID3 runs were used to incorporate the appropriate interaction terms in the MCA models developed to test hypotheses. In those cases where an individual independent variable was not included as an interaction effect in an MCA model, it was included as a main effect in accordance with the procedure established by Andrews et al (1973, p. 17-20). These main and interaction effects, as well as the major results for the four MCA models, are shown in Table 2. The F-ratios in Table 2 indicate how much of the variance of each dependent variable is accounted for by all the independent variables together.

TABLE 2

MCA MODEL RESULTS

While these results demonstrate favorable support for the notion that household characteristics are related to use of shopping strategies to reduce obligatory time, they do not indicate which of the terms in the MCA models yield the significant F-ratios. Table 3 provides more specific results for the main and interaction effects for each dependent variable. Means are provided for all classes present in each main effect and for selected classes in each interaction effect. The F-ratios reported in Table 3 are based on a statistical test that examines the proportion of the variance of the dependent variable that is explained by each independent variable while holding the other independent variables constant (Andrews et al. 1973, p. 34). These F-ratios are significant at the .05 level or less in 9 of the 20 model terms (45.0%). In the four instances where the interaction terms are included in the MCA runs, these F-ratios are significant at the .001 level. The importance of these household characteristics varies across the four dependent variables that were used in this study. As a result, discussion centers around the predictor variables and the success with which the descriptive model accurately specified their relationships to the strategies studied here.

Employment status of the female head of household. The data do not unequivocally support the expectation that the greater number of work hours are directly related to use of the four shopping strategies. Two of these dependent variables (trips by others and non-food purchases) generated non-significant F-ratios and the means of the other two variables are not precisely as one would expect. For example, the MCA model for convenience food purchases produced a significant interaction effect for female employment and marital status. As expected, the class mean for fully employed married women is higher than for married women who are either unemployed or work part time. The mean for married women employed full time (.152) is also higher than the mean for married women who are employed less than full time (.143). However, the class mean for married women who are employed part time is less than for those who are not employed, a finding counter to expectations. Furthermore, only minute differences exist between married women who are employed either full time or part time (.141), and married women who are not employed (.147). (Similar results were obtained for filler trips.) While findings such are these are counter to expectations, they have parallels in extant research (e.g., Anderson 1971; Strober and Weinberg 1977). It seems that the conclusions reached regarding the relationship between female employment and use of strategies to reduce obligatory time expenditures depends upon how one treats part-time employment, at least for the data reported here.

Employment status of the male head of household. As with female employment, male employment produced significant F-ratios for only two variables, filler trips and trips by others; for non-food and convenience food purchases the F-ratios were non-significant. However, unlike female employment, the mean scores for the categories of male employment were consistent with the expectation that full employment of the male head of household would create greater obligatory time and would result in use of obligatory-time reducers. However, if the male head of household were self-employed, those pressures might be less and if so should result in a lower frequency of use of shopping strategies which reduce obligatory time. Likewise, part-time employment or unemployment of the male head of household would result in a lesser need to use obligatory time reducers.

The findings for filler trips and trips by others confirm the hypotheses for male employment status. It is interesting that employment status of the male head of household has such an apparent impact upon a household's shopping strategies (for the variables studied here) while the employment status of the female head, a variable receiving considerable attention in the current literature, has an anomalous relationship to those same strategies.

Age of female head of household. Age was expected to be highly correlated with stage in the family life cycle since a younger woman would be more likely to be in the earlier . stages of the family life cycle. Being more likely to experience disequilibrium due to insufficient discretionary time, they would also be more likely to use the four shopping strategies studied here. In fact, age is the most successful of the independent variables in explaining variation in the dependent variables as indicated by significant F-ratios. As Table 3 indicates, age is a significant main or interaction term for three of the four dependent variables, convenience food purchase being the exception. Furthermore, the directionality is precisely as expected; households with younger women engage in a greater proportion of filler trips, purchase a higher proportion of non-food items, and are more likely to have others do the grocery shopping. Thus, if age of the female head of household is directly related to stage in the family life cycle, there is indirect evidence here in support of a piece of the time allocation equilibrium model.

TABLE 3

CLASS MEANS AND MCA RESULTS

Age distribution of household members other than male and female heads. Others has significant F-ratios for two variables (filler trips and non-food purchases), although the class means are inconsistent with expectations It was believed that households having infants needing care would be more likely to need to use strategies that reduce obligatory time and those without children the least likely to need such strategies. In fact, the data for filler trips are not at all consistent with prior expectations. Non-food purchases has class means that are somewhat more consistent with those expectations, but not totally so.

Household income. Income was one of the variables thought to facilitate the use of non-food and convenience food purchasing strategies. Unfortunately, neither F-ratio was significant. Unless the assumption that the use of these two strategies was not feasible unless one had the higher income to afford these more expensive purchases is incorrect, the time equilibrium model has not specified a variable contributing significantly to explanation of the variance in use of these strategies.

Education level of the female head of household. Education was included as a main effect only in the model for non-food purchases. A non-significant F-ratio fails to support the expectation that higher education would be associated with greater use of non-food Purchases.

Number of automobiles. The absence of automobiles was believed to be directly related to use of filler trips as the presence of a car would mean that major trips would be possible. The F-ratio is significant at the .10 level (and the means in the expected direction) providing weak support for the model.

Interaction variables. As Table 3 indicates, interaction effects provide considerable explanation for three of the four shopping strategies studied here. The time allocation model truly implies that it is the interactions among household characteristics that result in time disequilibrium. It is unlikely that any one of the variables discussed above could, by itself, so overwhelm the others to produce time disequilibrium needing resolution through the use of certain shopping strategies. The finding of four interaction variables, all of which are significant at or beyond the .001 level, provides evidence of the validity of the time allocation model.

CONCLUSIONS

The empirical results reported here demonstrate two positive features of the methodology. First, as a test of the model, the generally favorable results demonstrate the usefulness of the theory of time allocation suggesting that further empirical investigations could also benefit from examination of the model. Second, the data analysis methods allow for incorporation of interaction effects among predictor variables, an important feature of any operationalization of the time allocation process, although largely ignored in previous research (a problem discussed by Holman and Wilson 1980). Additionally, the research, combined with the results presented in Holman and Wilson (1982), provides insight into consumers' choices of shopping strategies and the variables that seem tn affect those choices.

REFERENCES

Abbott, M. and Ashenfelter, O. (1976), "Labour Supply, Commodity Demand and the Allocation of Time," Review of Economic Studies, 43, 389-411.

Anderson, W. T., Jr. (1971), "Identifying the Convenience-Oriented Consumer," Journal of Marketing Research, 8, 179-83.

Andrews, F. M., Morgan, J. N., Sonquist, J. A., and Klem, L. (1973), Multiple Classification Analysis (2nd ed.), Ann Arbor: Institute for Social Research, The University of Michigan .

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

Chapin, F. S., Jr. (1974), Human Activity Patterns in the City, New York: Wiley.

Hendrix, P. E., Kinnear, T. C., and Taylor, J. R. (1979), "The Allocation of Time by Consumers," in Advances in Consumer Research, Vol. 6, ed. W. L. Wilkie, Ann Arbor: Association for Consumer Research

Holman, R. H. and Wilson, R. D. (1980), "The Availability of Discretionary Time: Influences on Patterns of Consumer Behavior," in Advances in Consumer Research, Vol. 7, ed. J. C. Olson, Ann Arbor: Association for Consumer Research.

Holman, R. H. and Wilson, R. D. (1982), "Temporal Equilibrium as a Basis for Retail Shopping Behavior," Journal of Retailing, 58 (Spring), 58-81.

MacKay, D. B. (1973), "A Spectral Analysis of the Frequency of Supermarket Visits," Journal of Marketing Research, 10, 84-90.

Neter, J. and Wasserman, W. (1974), "Using AID and MCA to Analyze Marketing Data," in Combined Proceedings 1971 Spring and Fall Conference, ed. F. C. Allvine, Chicago: American Marketing Association.

Phlips, L. (1978), "The Demand for Leisure and Money," Econometrica, 46, 1025-43.

Prochaska, F. 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.

Robinson, J. P. (1977), How Americans Use Time, New York: Praeger.

Sonquist, J. A., Baker, E. L., and Morgan, J. N. (1973), Searching for Structure, (revised ed.), Ann Arbor: Survey Research Center, Institute for Social Research, The University of Michigan.

Stampfl, R. W. (1978), "The Consumer Life Cycle," The Journal of Consumer Affairs, 12, 209-19.

Strober, M. H. and Weinberg, C. B. (1977), "Working Wives and Major Family Expenditures," Journal of Consumer Research, 4. 141-7.

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

Vickery, C. (1977), "The Time-Poor: A New Look at Poverty," Journal of Human Resources, 12, 27-48.

Voss, J. and Blackwell, R. D. (1979), "The Role of Time Resources in Consumer Behavior," in Conceptual and Theoretical Developments in Marketing, eds. O. C. Ferrell, S. W. Brown, and C. W. Lamb. Jr.. Chicago: American Marketing Association

Wilson, R. D. (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, Iowa City: The University of Iowa.

Winer, B. J. (1971), Statistical Principles in Experimental Design, (2nd ed.), New York: McGraw-Hill.

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