An Examination of Consumer Grocery Store Choice: Considering the Attraction of Size and the Friction of Travel Time

When minor expenditures are contemplated by the consumer, the preference for nearby stores would be exaggerated while the preference for size would be reduced. In such a situation, travel costs are large in relation to purchase costs and weigh more heavily in decision making.

When major cash expenditures or a number of purchases are contemplated by the consumer, the reverse is true. The preference for size is exaggerated and the preference for nearby stores is reduced. The consumer can average the cost of the trip over more dollars or more items (Christaller, 1966; Downs, 1961; Bender, 1964). In addition, an outlet which does not allow the consumer to minimize travel costs when only one product is being purchased may allow minimization over the entire trip when the trip involves visits to multiple destinations.

RESEARCH PLAN

This paper reports on a test of some propositions involving consumer reactions to gravity model variables. Unlike most gravity model research, this study concentrates on a common managerial problem: understanding consumer response to the characteristics of a particular store. Most studies, in contrast, examine shopping centers or towns. Also, unlike most other studies, this research uses the individual consumer as a unit of analysis rather than an entire neighborhood.' Individual level analysis has been recommended by several authors (Bucklin, 1967; and Mackay, 1970). A final distinction is that this study uses measures of actual behavior as opposed to reports of behavior from survey questionnaires.

Hypotheses

The following general questions are addressed: Do consumers react to travel friction as if it is a cost to be overcome and to size as a store benefit? Do consumers act as if the importance of these costs and benefits changes with the sizes of payoffs sought on different types of trips? Six specific hypotheses are tested.

The first two hypotheses are fundamental in the basic gravity model formulation examined in the context of grocery shopping.

Hypothesis I: Consumers will make more visits to stores nearby their homes than they will make to more distant stores.

Hypothesis II: Consumers will make more visits to larger stores than they will make to smaller stores.

The second pair of hypotheses explore single-and multiple-purpose trips. A single-purpose grocery shopping trip is made by a consumer when the grocery store is the only destination contemplated and when products are purchased only from the grocery store. A multiple-purpose trip is made by a consumer when some other destination is contemplated in addition to the grocery store.

On multiple-purpose trips, one expectation is that the consumer should be willing to incur greater travel costs. Another expectation is that the consumer should desire to concentrate purchases at a single large center rather than traveling to several different places.

Hypothesis III: Consumers will patronize stores further from their homes on multiple-purpose grocery shopping trips than the centers they will patronize on single-purpose trips.

Hypothesis IV: Consumers will patronize larger shopping centers on multiple-purpose grocery shopping trips than the centers they will patronize on single-purpose trips.

A final set of hypotheses explore convenience and major grocery shopping trips. A "convenience" trip is operationally defined to consist of the purchase of four or fewer items, which are not the specialty of the store, at a cost of $5.00 or less. Other trips were labeled as "major" if the consumer spent eighteen percent or more of her grocery budget for a month. These cut-off points are arbitrary, but they capture the essence of the distinction being made. Of course, a number of trips did not fall into either category.

Hypothesis V: Consumers will patronize stores closer to their homes on convenience grocery shopping trips than the stores they will patronize on major trips.

Hypothesis VI: Consumers will patronize smaller grocery stores on convenience grocery shopping trips than the stores they will patronize on major trips.

Data

The basic data used to test these hypotheses were taken from a panel of 10! consumers living in a number of suburbs northwest of Chicago. The area was characterized by consistently high levels of store availability and accessibility. It also provided an array of shopping alternatives: "mom and pop" stores, convenience stores, small older supermarkets, and new larger supermarkets.

The panel consisted of white, middle class, suburban females who used the automobile to do their shopping. These women made a total of 992 visits to local grocery stores over a period of a month. Their trips provide the dependent measure in the analysis. Independent measures include: store sizes, shopping center sizes, and travel distances to stores.

Data Analysis

Three methods of data analysis were attempted. Ordinary least squares regression analysis was employed to test

Hypotheses I and II. The basic equation to be estimated was simply:

The independent variable in Hypothesis I, D_i, represents the travel distance separating consumers' homes from patronized stores. Distance is represented by 20 quarter mile intervals (i = 1, . . . , 20). A five mile boundary was chosen because only four of the original 992 trips were made to stores beyond that point.

For any aggregate of shops, it is not too extreme to assume an area to be served. No doubt a native of Detroit might some day buy an article in Birmingham, Alabama, or even in Hong Kong, but it is doubtful whether a retailer of interest to central place analysis takes this into consideration. It seems fair to assume that the territorial ambitions of the most aggressive of hucksters has a limit. To facilitate analysis, it will be assumed here that a local area ends abruptly at its edge, and beyond stretches the void (Curry, 1967, p. 221).

The dependent variable in Hypothesis I, P_i - O_i, is a combination of two variables. The behavioral variable of interest, Pi, is the proportion of total visits made to a store within a particular distance interval. The geometry of the spatial choice situation requires the inclusion of a structural variable, O_i, to correct for the "pattern of opportunity". O_i is an index showing that the proportion of the total area in the five mile range which falls within a particular interval. Each successive zone radiating outward from the consumer's home has more area and a higher probability of stores being located there. If distance were not a friction variable, then, more trips would be made to more distant stores simply because of availability.

The independent variable in Hypothesis II, S_j, is the size of the grocery store measured in square feet of floor space. Size was represented by 10 intervals containing 5,000 square feet. The dependent variable, P_j, is the proportion of total visits made to a store in a particular size interval.

Two additional methods were used to test Hypotheses III through VI. The first was a simple test of mean differences between types of trips. The analysis was based on the following hypotheses (Hays and Winkler, 1971, p. 428).

H₀: X_i - X_j = 0      H₁: X_i - X_j > 0

where:

X = average value

i = consumers making one type of trip

j = the same consumers making the other type of trip

The final method for testing Hypotheses III through VI was regression analysis using indicator variables (Neter and Wasserman, 1974, p. 304). An attempt was made to identify separate linear distance decay functions for the different trip types.

Comparative analyses are restricted to those persons who happened to engage in both types of behavior. So the sample sizes vary among tests.

RESULTS

Patronage displays a pattern of "distance decay" envisioned in Hypothesis I, that is, more consumer visits were made to stores in nearby distance intervals than to more distant stores. The pattern is basically negative monotonic as expected. A deviation from the trend appears in the interval from 0 to 1/4 miles. Because of zoning restrictions in the study area, few homes are located this close to stores.

The following regression equation estimates the distance decay function:

P_i - O_i = .12765 - .01218 (D_i)       r² = .89

                             (.001)

Where:

P_i = (number of visits in distance interval I) / (Total number of visits)

D_i = travel distance separating consumers' homes from patronized stores

O_i = (the area in distance interval I) / (total area within 5 miles of consumers' homes)

Considering store size in Hypothesis II, the results are not as straightforward. The pattern of patronage is approximately bell shaped over different store size intervals, so linear regression is not appropriate. The pattern is certainly not monotonically increasing as expected. Thus, the data do not indicate that consumers respond to store size as a general benefit.

Consumers who made both multiple-and single-purpose trips acted as expected in Hypotheses III and IV. The results are shown in Table 1. They chose nearer stores when they were buying only groceries and more distant stores when also making non-grocery store purchases. The difference amounts to about a ten percent increase in distance beyond the single purpose stores. However, in regression analysis, it is not possible to isolate two separate linear distance decay functions using indicator variables to represent different trip types. Consumers also chose grocery stores in larger shopping centers when making multiple purpose trips than when making single purpose trips. The difference amounts to about a 96 percent increase in size over single purpose shopping centers.

Consumers who made both major and convenience grocery shopping trips also acted as expected in Hypotheses V and VI. The results are shown in Table 2. They chose more distant stores when the amount of purchase was large and closer stores when purchase amounts were insignificant. The increase in distance is about 27 percent more for major trips than for convenience trips. Again it is not possible to isolate separate linear distance decay functions for the two different trip types. Consumers also chose larger stores on major trips than on convenience trips. The increase of major over convenience amounts to about 40 percent.

TABLE 1

DIFFERENCES BETWEEN MULTIPLE AND SINGLE PURPOSE GROCERY SHOPPING TRIPS: TRAVEL DISTANCES TO AND SIZES OF SHOPPING CENTERS PATRONIZED

TABLE 2

DIFFERENCES BETWEEN CONVENIENCE AND MAJOR GROCERY SHOPPING TRIPS: TRAVEL DISTANCES TO AND SIZES OF GROCERY STORES PATRONIZED

DISCUSSION

The statistical tests employed in this study produce mixed although moderately successful results. Hypotheses I, IV, and VI, are supported completely. Hypotheses III and V receive qualified support. They pass the t test, but the trends envisioned in the hypotheses can not be described by separable linear functions in regression analysis.

Only Hypothesis II receives zero support. "Average" sized stores received more patronage than either larger or smaller stores. The pattern looks as if it could have been generated from a random selection process or as if the most popular stores were average size. Several explanations are possible: First, the results observed may reflect the actual pattern of available stores rather than a preference function for store sizes. Second, the preference function for store size may be non-monotonic as Baumol and Ide (1956) suggest. Size to them is thought to be attractive up to a certain point. Beyond that point increases in size mean more shopping cost for the consumer rather than better selection. Finally, Bucklin's (1967) admonishment about gravity model applications to individual stores may apply:

Doubt that mass will be adequate is brought to mind because of difficulties experienced by researchers attempting to evaluate inter-store trading areas. At this level of disaggregation, the image content of shopping utility evidently becomes the controlling factor (p.38).

However, this final assertion seems inconsistent with the favorable result for Hypothesis VI which indicates that consumers were reacting to store size as a source of utility.

The results of this study lend credence to the general location theory postulates about consumer behavior. Consumers do appear to react to distance and store size in choosing stores. In addition, the present test is unusual. It demonstrates that location theory can be tested using individual stores and individual consumers as units of analysis. Although Huff (1966) suggests that both of these uses are possible, they are rarely seen. The test is also unique in that it uses actual behavior as a dependent variable rather than behavior reports from surveys.

The study would have certainly been improved by a larger sample size. Then both the multiple-single and the major-convenience distinctions could have been investigated simultaneously. The expectation should be that a single-purpose, convenience trip should encourage patronage at the nearest store of any size.

From a managerial perspective, an important implication lies in the realm of trading areas. Smaller trading areas will be controlled by stores which cater to convenience and single-purpose shoppers, and larger trading areas will be enjoyed by stores which cater to major trip and multiple-purpose shoppers.

What happens if the gasoline supply is restricted in the future? Two consumer options are obvious: a reduction in automobile travel or a switch to alternate modes of travel (Becker, Brown, and Schary, 1976). In the first case, carefully planned multiple-purpose trips or a small number of major trips would be in order for the consumer. The effect could be to increase the trading areas of large stores and large shopping centers.

Switching to non automotive modes of travel (bicycle, walking, and mass transit, for example) generally means that smaller amounts can be brought home from a shopping trip. Furthermore, alternatives such as walking or biking greatly increase the friction of distance. The effect could be to decrease the trading areas of stores so that the economies of scale from large operations would no longer be possible. In the case of either consumer transportation response, the effect on the pattern of retail trade can be expected to be dramatic.

REFERENCES

W. J. Baumol and W. A. Ide, "Variety in Retailing," Management Science, 3(October, 1956), 93-101.

Boris W. Becker, Daniel J. Brown and Philip B. Schary, "Behavior of Car Owners During the Gasoline Shortage," Traffic Quarterly, 30(July, 1976), 496-483.

Wesley C. Bender, "Consumer Purchase Costs -- Do Retailers Recognize Them?" Journal of Retailing, 40(Spring, 1964), 1-8 and 52.

Louis P. Bucklin, "The Concept of Mass in Interurban Shopping," Journal of Marketing, 31(October, 1967), 38-42.

Louis P. Bucklin, Shopping Patterns in an Urban Area (Berkeley: Institute of Business and Economic Research, University of California, 1967).

Louis P. Bucklin, "Retail Gravity Models and Consumer Choice: A Theoretical and Empirical Critique," Economic Geography, 47(October, 1971), 489-497.

Walter Christaller, The Central Places of Southern Germany (Englewood Cliffs, N. J.: Prentice Hall, 1966).

Leslie Curry, "Central Places in the Random Spatial Economy," Journal of Regional Science, 7(Winter, 1967), 217-237.

Anthony Downs, "A Theory of Consumer Efficiency," Journal of Retailing, 37(Spring, 1961), 6-12.

William L. Hays and Robert L. Winkler, Statistics: Probability, Inference and Decision (New York: Holt, Rinehart, and Winston, Inc., 1971).

David L. Huff, "A Probabilistic Analysis of Consumer Spatial Behavior," Proceedings of the Fall and Winter Conference of the American Marketing Association, 1962, 443-461.

David L. Huff, "A Programmed Solution for Approximating an Optimum Retail Location," Land Economics, 42(August, 1966), 293-303.

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

David B. MacKay, Consumer Movement and Store Location Analysis, Doctoral Dissertation, Northwestern University (Evanston, Ill., 1970).

Masao Nakanishi and Lee G. Cooper, "Parameter Estimation for a Multiplicative Competitive Interaction Model -- Least Squares Approach," Journal of Marketing Research, 11 (August, 1974), 303-311.

John Neter and William Wasserman, Applied Linear Statistical Models (Homewood, Illinois: Richard D. Irwin, 1974).

William J. Reilly, The Law of Retail Gravitation (Austin: University of Texas, 1931).

Thomas J. Stanley and Murphy A. Sewall, "Image Inputs to a Probabilistic Model: Predicting Retail Potential," Journal of Marketing, 40(July, 1976), 48-53.

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