7 ,i,,.__T ..n.‘— ”(a p __._ ,_.V. ,h ._,A. .. - « ~-.»«— \V--..-.‘..-w~:~,m*~ AN ANALYSIS OF THE PURCHASE CLUSTERING PATTERNS OF FOOD SHOPPERS Thesis for the Degree of Ph. D. MICHIGAN STATE UNIVERSTTY VELAGAPUDE KANTA PRASAD 1970 "Ll!IHIILILIIMIJUUILIIJMMIIZII”WIN “yuan“? ‘ w Michigan 5m UniVCT-‘iw (9 CF" This is to certify that the thesis entitled An Analysis of The Purchase Clustering Pa++erns of Food Shoppers presented by Velagapudi Kanta Prasad has been accepted towards fulfillment of the requirements for Ph.D. . Marketing ____.____ degree m— /,/‘:“‘7- v. 6 , ‘ ”Major professor‘ K 9 70 Date 5/2 / 0-169 ’4‘ ”‘7 k *z- ‘2»... t y emanate av ‘3 1 "MG & SONS’ 800K BINDERY INC. :- LIBRARY BINDEHS ABSTRACT AN ANALYSIS OF THE PURCHASE CLUSTERING PATTERNS OF FOOD SHOPPERS BY Velagapudi Kanta Prasad In an environment of increasing competitive pres- sures and rising costs, the trend in the recent years toward a greater incidence of multiple-store food shOpping among consumers has been a matter of concern to the food retail industry. Multiple—store shOpping by consumers, taking into consideration the number of food stores patronized as well as the dollar expenditures spread among these stores was referred to in the study as 'food purchase clustering' of shoppers and was the primary focus of the research. The purpose of the research was to investigate if shOppers who exhibited different degrees of food purchase clustering could be identified in terms of selected char- acteristics of the shOppers. The characteristics examined were: (1) socio-economic and demographic variables, (2) Selected food purchasing characteristics of shOppers, and (3) role-related self-perceptions of housewives. The research also examined if there exists a significant re— lationship between shoppers' perceptions of similarity Velagapudi Kanta Prasad among the food stores patronized by them and the patterns of clustering of their food purchases among these stores. The research was conducted in the city of Lansing, Michigan. Data were collected through self-administered questionnaires mailed to the homemakers of one thousand families who were selected according to a multi-stage strat- ified sampling procedure. The findings reported in the research were based on a total of 335 usable questionnaires returned by the sample families. A measure of the degree of food purchase clustering exhibited by families was deve10ped by the research. The data were analyzed using apprOpriate statistical methods. The major findings of the research were as follows: 1. The predictive efficacy of socio-economic and demographic variables in explaining differences in the food purchase clustering patterns of sh0ppers was very low. However, two of the variables, the stage in the family life cycle and multiple-automobile availability, were found to be significantly related to the extent of food purchase cluster— ing of families. Families in the earlier stages of the life cycle clustered their food purchases to a relatively greater degree than those in the other stages of the life cycle. 2. .Families who clustered their food purchases to a relatively greater extent were observed to have generally lower food budgets, do food shOpping less frequently and Spend lesser amounts of in-store grocery sh0pping time than Velagapudi Kanta-Prasad others. The degree of food purchase clustering was also found to be significantly related to the extent of multi- purpose food shOpping on the part of the families. 3. Role-perception characteristics of the home- makers were found to be poor indicators of food shOppers' patterns of purchase clustering among stores. 4. ShOppers' comparative perceptions of the food stores they patronized with reSpect to prices and quality of meats were significantly related to the patterns of cluster- ing of their food purchases among these stores. The research has a number of implications for super— market management and consumer behavior research. 1. Supermarket managements may achieve a more favor— able 'customer loyalty mix' by carefully assessing the needs and wants of shOpper families who are in the earlier stages of the family life cycle and suitably adjusting the merchan- dising and promotional efforts to increase the patronage of this shOpper segment. 2. The research calls for more careful evaluations of decisions to locate supermarkets in shOpping centers. It suggests that a supermarket in order to be located in a Shop- ping center should be first justifiable as a good food loca- tion with respect to the consumer pOpulation in the relevant trading area who treat food shOpping as a single-purpose activity. Velagapudi Kanta Prasad 3. The generally low predictive efficacy of the major groups of variables included in the research indicates the need for more search for important factors which influ- ence food purchase clustering patterns of shOppers. The findings appear to generally support a growing realization among researchers that investigations of determinants of purchase behavior, to be fruitful, should consider character- istics that are idiosyncratic to both the customer and the product (or the purchase situation) and not to the customer alone as in the case of socio-economic variables or role- perception characteristics. AN ANALYSIS OF THE PURCHASE CLUSTERING PATTERNS OF FOOD SHOPPERS BY Velagapudi Kanta Prasad A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Marketing and Transportation Administration 1970 c- we” /—:;ov7/ ACKNOWLEDGMENTS The author wishes to express his sincere appreciation to the following persons who have been instrumental in the completion of this research effort: Dr. William J. E. Crissy, Professor of Marketing and Transportation Administration, Michigan State University, as Chairman of the Committee which guided the research, has provided invaluable direction and motivation from the incep- tion to the completion of the research. His ideas, comments and suggestions have contributed substantially_to the refine- ment of the research effort, and his generosity in time and effort is most gratefully appreciated. Dr. Crissy's dynamism and concern for students will always be sources of inspira- tion for the author. Dr. John W. Allen, Director, Food Marketing Program, Department of Marketing and Transportation.Administration, Michigan State University, as a member of the Committee has given his time and encouragement freely throughout all stages of the work. .Dr. Allen's comments and suggestions have been substantially helpful in the conduct of the research. Dr. Richard F. Gonzalez, Professor of.Management, Michigan State University, as a member of the Committee, has ii greatly aided the research effort through his suggestions and constant encouragement. As a principal investigator of an earlier research project for which the author was one of the investigators, Dr. Gonzalez helped the author to gain valuable experience in mathematical model building in the business field. His encouragement throughout the author's doctoral program is sincerely appreciated. Dr. Bernard J. LaLonde, James R. Riley, Professor of Marketing and Logistics, The Ohio State University, formerly Professor of Marketing and Transportation Adminis- tration, Michigan State University, gave generously his time and advice which were of substantial help in the formulation of a prOposal for the present research. His time and encour- agement are much appreciated. Financial support for the research study was aided substantially by a research grant from the Food Marketing Program, Department of Marketing and Transportation Admin- istration. Finally, the author wishes to express his apprecia- tion of the help rendered by Mary Kennedy in the organization of the mail survey. iii TABLE OF CONTENTS Page ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . ii LIST OF TABLES . . . . . . . . . . . . . . . . . . . . Vii LIST OF FIGURES . . . . . . . . . . . . . . . . . . . ix Chapter I. INTRODUCTION . . . . . . . . . . . . . . . . . 1 Nature of the Problem . . . . . . . . . . . 1 Statement of the Problem . . . . . . . . . . 4 Hypotheses . . . . . . . . . . . 5 Research Design and Methodology . . . . . . 9 Limitations of the Study . . . . . . . . 10 Potential Contributions of the Research . . 11 Organization . . . . . . . . . . . . . . . . 14 II. A REVIEW OF CONSUMER LOYALTY'RESEARCH . . . . 15 Consumer Loyalty--The Early Studies . . . . 16 Consumer Loyalty--Some Theoretical Constructs . . . . . . . . . . . . . . . . 19 Learning Theory . . . . . . . . . . . . 20 Image Congruence . . . . . . . . . . . . 24 Risk Taking Theory . . . . . . . . . . . 28 Group Influence . . . . . . . . . . . 29 Brand Loyalty and.Market Segmentation Research . . . . . . . . . . 33 Identifiability of Brand Loyal Customer Segments . . . . . . . . . . 35 'Differentiability of Purchase Characteristics and Elasticities of Promotion of Brand Loyal Customers . . . . . . . . . . . . 38 Empirical Research on Store Loyalty . . . . 40 Other Relevant Research . . . . . . . . . . 46 Summary . . . . . . . . . . . . . . . . . . 47 iv Chapter III. IV. RESEARCH DESIGN . . . . . . . . . . . . . . Identification of Research Variables . . . Independent Variables . . . . . . . . Dependent Variable . . . . . . . . Additional Analysis and Relevant Variables . . . . . . . . . . . . . Sample Design . . . . . . . . . . . . . The Sampling Frame . . . . . . . . . Selection of Sample Households . . Data Collection . . . . . . . . . . . Analysis of the Data . . . . . . . . . . . Data Preparation . . . . . . . . . . Computer Programs for Statistical Analysis . . . . . . . . . . . . . . PRESENTATION OF FINDINGS . . . . . . . . . . Variations in the Food Purchase Clustering Patterns of the Sample Families . . . . . . . . . . . . . . . Socio-Economic and Demographic Variables . Family Income . . . . . . . . . . . Educational Level of the Homemaker . . Employment Status of the Homemaker . . Occupational Status of the Household Head . . . . . . . . . Multiple-Automobile Availability . . . Stage of the Family Life Cycle . . . Family Size . . . . . . . . . . . . Age of the Homemaker . . . . . . . . Number of Pre-School Age Children . . Family Income Versus Other Socio-Economic and Demographic Variables . . . . . . . Combined Predictive Efficacy of Socio- -Economic and Demographic Variables . . Purchasing Characteristics . . . . . . . . Total Grocery Expenditures . . . . . . Frequency of Grocery Shepping . . . Extent of Multi-Purpose Food ShOpping In-Store Grocery ShOpping Time . . . . Combined Predictive Efficacy of Purchasing Characteristics . . . . . . . . . . . . RoleéRelated Self—Perceptions . . . . . . Degree of Food Purchase Clustering and Perceptual Similarity of Stores . . . . Summary . . . .... . . . . . . . . . Page 49 50 50 52 56 57 57 58 59 61 61 62 65 66 68 69 71 71 74 76 78 81 81 84 86 88 89 90. 92 94 96 98 100 101 107 Chapter Page V. SUMMARY AND CONCLUSIONS . . . . . . . . . . . 113 General Summary of the Research Study . . . 113 Evaluation of the Research Hypotheses and Presentation of the Major Conclusions . . . . . . . . . . . . . . . 116 Socio-Economic and Demographic Characteristics . . . . . . . . . . . 116 Other Food Purchasing Characteristics . 119 Role-Related Self-Perceptions of Homemakers . . . . . . . . . . . . . . 121 Store Perceptions . . . . . . . . . . . 121 Implications of the Research Findings . . . 122 Implications of the Research for Supermarket Management . . . . . . . . 122 Customer Loyalty'Mix . . . . . . . . 123 Supermarket Location Decisions . . . 126 Store Images and Store Loyalty . . . 128 Implications of the Research for Consumer Behavior Research . . . . . . 129 Suggested Areas for Further Research . . . . 130 Appendix A. COVER LETTER AND RESEARCH QUESTIONNAIRE . . . 133 B. SOCIO-ECONOMIC AND DEMOGRAPHIC COMPOSITION OF THE SAMPLE FAMILIES . . . . . . . . . . . 138 BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . 140 vi Table 1-1. LIST OF TABLES Extent of Multiple-Store Shopping for Food . . . . . . . . . . . . . . . . . . Variation in Purchase Clustering Patterns of the Sample Families . . . . . . . . . Degree of Purchase Clustering by Family Income 0 O C O O O O O O O O O O O O O 0 Degree of Purchase Clustering by Educa- tional Level of the Homemaker . . . . . . Degree of Purchase Clustering by Employment Status of the Homemaker . . . Degree of Purchase Clustering by Occupa- tional Status of HouSehold Head . . . . . Relationship Between Degree of Purchase Clustering and Multiple-Automobile Availability . . . . . . . . . . . . . . Degree of Purchase Clustering by-Stage in the Family Life Cycle. . . . . . . . . Degree of Purchase Clustering by Size of the Family . . . . . . . . . . . . . . Degree of Purchase Clustering by Age Group of the Homemaker . . . . . . . . . Degree of Purchase Clustering by Number of Preschool Age Children in the Family . Analysis of Variance for Testing the Significance of Socio—Economic and Demographic Variables as a Group . . . . Degree of Purchase Clustering by Total Grocery EXpenditure . . . . . . . . . . . vii Page 67 7O 72 73 75 77 79 82 83 85 88 91 Table 4-13. Page Degree of Purchase Clustering by Frequency of Grocery Shopping . . . . . . . 93 Relationship Between Degree of Purchase Clustering and Extent of Multi-Purpose Food ShOpping . . . . . . . . . . . . . . . 95 Degree of Purchase Clustering by In-Store Grocery ShOpping Time . . . . . . . . . . . 97 Analysis of Variance for Testing the Significance of Purchasing Character— istics as a Group . . . . . . . . . . . . . 99 Analysis of Variance for Testing the Significance of Role—Related Perceptions of Housewives . . . . . . . . . . . . . . . 101 .Degree of Purchase Clustering and Store Price and Proximity Perceptions . . . . . . .103 Degree of Purchase Clustering and Store Proximity Perceptions . . . . . . . . . . . 105 Degree of Purchase Clustering and Store Price Perceptions . . . . . . . . . . . . . 106 Degree of Purchase Clustering and Perceptions of Quality of Meats . . . . . . 108 viii Figure 2-1. LIST OF FIGURES Page Howard-Sheth paradigm of consumer brand loyalty . . . . . . . . . . . . . . . . . . . 21 Graphical illustration of the entrOpy measure of purchase clustering when the shOpper patronized only two stores for her food purchases . . . . . . . . . . . . . 55 ix CHAPTER I INTRODUCTION Nature of the Problem In a marketing oriented economy, knowledge of rele- vant patterns of customer behavior is essential to the success of firms. Over the past decades the food retail industry in the United States adapted itself remarkably to shifting consumer needs and purchase habits. However, the need for studies that provide better insights into food shOppers' purchase behavior and discern and analyze signif- icant trends of change in food shOpping behavior is a con— tinuous one. Super market industry has been under a continuing profit squeeze over the recent past years. According to Progressive Grocer, average net Operating profit of food chains had reached a new low Of 0.49 percent1 during 1967-68. Increasing pressures of competition and rising costs are commonly recognized as some of the contributing factors. 1"Thirty-Sixth Annual Report of the Grocery Industry," Progressive Grocer, April, 1969, p. 69. In part, the rising competitive pressures are the resultant of a significant growth in the number of supermarkets1 whose market share of grocery business has reached the mark of 79 percent2 and also in the number of convenience stores.3 Viewed in the context of the competitive environment in which stores have to strive for customer patronage, it is of great concern to grocery store managements that there has been a significant trend toward multiple-store shopping by consumers to fulfill their food buying needs. Progressive Grocer observed, ". . . customers free to pick and choose among many markets of similar nature, have been Spreading their purchases among two, three or even more supermarkets. Store loyalty, many Operators have to come to realize, has 4 Data collected by Burgoyne Index sunk to alarming lows." Inc., through national surveys of food shOppers indicate the trends in the extent of multiple-store shOpping of supermarket shOppers (Table 1-1). l1969 Supermarket Sales Manual--Chain Store Age, Vol. 45, Number 7A (Mid-July, 1969), p. 8. 21bid. 31bid., p. 23. 4"Food Retailing 1975: A Look Into the Future," Progressive Grocer, April, 1966, p. 153. TABLE 1-1 EXTENT OF MULTIPLE-STORE SHOPPING FOR FOOD Percentage of Supermarket Shoppers Patronizing 1954 1961 1963 1965 1967 (%) (%) (%) (%) (%) One supermarket exclusively 41 29 25 17 16 More than one supermarket 59 71 75 83 84 Source: Adapted from "The Fourteenth Annual Study of Super- market ShOppers" (Cincinnati, Ohio: Burgoyne Index, Inc., 1967). Although the above data give an idea of multiple- store food shOpping in terms of the number of stores patron- ized by consumers, they do not indicate the extent to which food expenditures of consumers in dollar terms are spread among different stores. Multiple-store food shOpping taking into consideration both the number of stores and expenditure spread among stores is referred to in the present study as the "purchase clustering behavior" of food shOppers, and is the primary focus of the research. The purpose of the research study is to investigate if different degrees of food purchase clustering can be identified by selected characteristics of shOppers. Statement of the Problem Purchase behavior is the resultant of a complex interaction of factors, some pertaining to the consumer and some pertaining to the object of the choice behavior. However, the research study does not attempt to pinpoint the motives and causal factors behind purchase clustering behavior of consumers although some inferences of such nature could possibley be drawn from the findings of the study. Spreading food purchases among several stores can be eXpected to involve some additional effort on the part of food shOppers in terms of travel, familiarization with mer— chandise layout, information search regarding prices, deals and other factors. The additional effort could be considered as part of the shOpper's secondary purchase costs which in the Shopper's perception are more than compensated for by the benefits derived from multiple-store shopping either in terms of matching her food needs more Specifically or in terms of monetary savings. The primary purpose of the research is to' examine if the (l) socio-economic and demographic variables, (2) the purchasing characteristics and (3) the role-related self-perceptions of homemakers can distinguish between food shOppers of differing degrees of purchase clustering. The research is guided by the following questions: 1. Do families exhibit significant differences in their food purchase clustering patterns? Is family income the variable among the socio— economic and demographic characteristics of shOppers most closely related to food purchase clustering patterns? Which of the other socio— economic and demographic variables are significantly related to the degree of food purchase clustering of families? Are socio-economic and demographic variables, considered as a group, significant in eXplaining differences in the food purchase clustering patterns of families? Which variable among the selected purchasing characteristics of the families have significant associations with their degree of food purchase clustering? Do the selected purchasing character- istics considered as a group have significant in- fluence on the food purchase clustering patterns of families? Do differential self-perceptions of homemakers with respect to selected role-related activities explain significantly differences in their food purchase clustering patterns? Is there a significant relationship between shOppers' comparative perceptions of the food stores they patronize and the patterns of clus- tering of their purchases among these stores? Hypotheses The questions regarding correlates and patterns of food purchase clustering behavior of families have been formulated in terms of the following testable hypotheses and subhypotheses. They have been stated in the positive format only for the sake of convenience. I. Socio-Economic Status Variables A. Family Income: The degree of food purchase cluster- ing1 of a family is significantly related to the total income of the family. Educational Level of theggomemaker: The degree of food purchase clustering of a family is signifi- cantly related to the educational level of the homemaker. Employmentggtatus of the Homemaker: The degree of food purchase clustering of a family is signif- icantly related to the employment status of the homemaker. Occupational Status of the Hogsehold Head: The degree of food purchase clustering of a family is significantly related to the occupational status of the household head. Multiple-Automobile Ownership: The degree of food purchase clustering of a family is significantly PP- 1For a definition of the term, see Chapter III, 52-56. related to the number of automobiles available to its members. II. Demographic Status Variables A. Stage in the Family Life Cycle: The degree of food purchase clustering of a family is significantly related to its stage in the family life cycle. B. Family Size: The degree of food puchase clustering of a family is significantly related to its size. C. Age o§,the Homemaker: The degree of food purchase clustering of a family is significantly related to the age of the homemaker. D. Number of Pre-School Age Childrgg; The degree of food purchase clustering of a family is signifi- cantly related to the number of pre-school age children in the family. III. Socio-Economic and Demographic Status Variableg A. .Family income is the most significant variable among the selected socio-economic and demographic status variables in explaining differences in the degree of food purchase clustering of families. B. Socio-economic and demographic status variables as a group are significant in explaining differences in the degree of fOOd purchaSe clustering of families. IV. Purchasing Characteristics 1.A. Total Grocery Expenditure: The degree of food purchase clustering of a family is significantly related to its level of grocery expenditures. 1.D. VI. Frequency of Grocery Shogping: The degree of food purchase clustering of a family is significantly related to the frequency of grocery shOpping of the homemaker. Extent of Multi—Pgrpgse Food ShOpping:l The degree of food purchase clustering of a family is signifi- cantly related to the extent of multi-purpose food shOpping of the homemaker. In-Store ShOppingTime: The degree of food purchase clustering of a family is significantly related to the average amount of weekly sh0pping time spent by the homemaker in grocery stores. Purchasing characteristics as a group are signifi- cant in explaining differences in the degree of food purchase clustering of families. Role—Related Self-Perceptions: Differences in self- perceptions of homemakers with respect to selected role- related activities explain significantly differences in their degrees of food purchase clustering. Among multiple-store shOppers of food, those who per- ceive their first and second choice stores as similar in terms of geographic proximity and price image have significantly lower degrees of food purchase clustering with respect to the two stores than other sh0ppers. p. 52. 1For a definition of the term, see Chapter III, Research Design and Methodology Mailed questionnaires were used to collect data on food shOpping, household socio-economic and demographic characteristics and role-related self-perceptions of 335 homemakers in the city of Lansing, Michigan. A multi-stage sampling procedure was employed to select the subjects. Using 1960 census data, census tracts were stratified into five groups on the basis of the median incomes of the tracts. City blocks in each stratum were enumerated and a prespec- ified number of blocks were randomly selected from each stratum. Using the 1969 edition of R. L. Polk's City Direc- tory - Lansing, Michigan,1 systematic random samples of households were chosen from each city block. Questionnaires were mailed on November 15, 1969. Follow up letters were mailed two weeks later requesting c00peration from nonérespondentSr; Any questions that the respondents might have had in filling the questionnaires were answered over telephOne. Responses sent back over a period of four weeks after the questionnaires were mailed, have been used as the data base for the study. The data pertaining to the usable questionnaires were coded and transferred to punch cards for tabulation and statistical testing of the research hypotheses. 1R. L. Polk, Polk's Lansing (Ingham County, Mich.) City Directory (Detroit, Michigan: R. L. Polk and Company, 1969). 10 Limitations of the Study The results of the research are subject to the following limitations: 1. The investigation was confined to one metrOpolitan area, Lansing, Michigan. Hence the problem of gen- eralizing from the results arises. Replication in another location may be needed to increase the degree of confidence in the conclusions of the study. -Due to the high costs involved, efforts to conduct a longitudinal study had to be abandoned. Purchase data were collected on the basis of recall on the part of homemakers and may be considered accurate only to that extent. Purchase data based on con- sumer diaries over an extended period, leaving aside cost considerations, could be eXpected to provide a more reliable data base for the study. The dependent variable used in the study, the degree of food purchase clustering, is a time-averaged mea— sure of purchase behavior rather than one that takes into consideration the time sequence of successive food purchases. This limitation should be kept in mind in interpreting the results of the research. 11 Potential Contributions Of the Research A vast amount of research in purchase behavior in recent years has been concentrated in the area Of brand purchase behavior and brand loyalty phenomena. A similar emphasis on research in stOre choice behavior and store loyalty has been lacking. Some empirical studies have been concerned with customer store loyalty but have dealt only with product-specific store loyalty rather than store loy- alty based on aggregate food purchases Of consumers. The present study contributes toward filling the above mentioned gap in purchase behavior research to some extent. The primary contribution of the present research is to develop a body of knowledge about the characteristics Of food shOppers that may effectively discriminate between those with high and low degrees Of food purchase clustering among stores. The research affords an Opportunity to examine the efficacy of personal attributes in explaining differences among consumers in an important aspect of purchase behavior and thus should be of significance to those engaged in market segmentation research. Also, the 'entrOpy measure' used in the study to measure the dependent variable--food purchase clustering among stores, extends the existing store loyalty measures in that it takes into account both the number of stores visited by the consumer as well as the prOportions of total food expenditure spent in each of the stores. 12 Another contribution Of the research is to provide an understanding Of the significance of role-related self- perceptions of homemakers in explaining differences in their store loyalty patterns in relation to food shOpping. The role Of household purchasing agent by a wife has been Often referred to in the marketing literature.l However, the housewife performs in a number of other interacting roles which influence her buying behavior as the household pur- chasing agent. Some studies2 have emphasized the influence Of role-perceptions of hOusewives on their food purchasing decisions. No attempt has been made in previous research, however, to examine if role—perceptions of homemakers are significantly related to their store loyalty patterns in relation to food shOpping. The present research makes a beginning in this direction. The role-perception character- istics of homemakers may prove to be important considera- tions in future market research. The present study attempts to extend the existing empirical research that relates food shOpping behavior to trip purpose (i.e., single-purpose versus multi-purpose). lWroe Alderson, Marketing Behaviorggpd Executive Action (Homewood, Illinois: Richard Irwin, Inc., 1957), p. 179. 2Howard Trier, Henry Clay and James Shaffer, uDifferences in Food Buying Attitudes Of Housewives," Journal Of Marketing, VOl. 25 (July, 1960), p. 67; and Louis P. Bucklin, "Consumer Search, Role Enactment and Marketing Efficiency," The Journal of Business, Vol. 42 (October, 1969), p. 435. 13 Past research1 investigated the relationship between trip purpose and the average distance the consumer is willing to travel for fOOd purchases. The present study attempts to relate the extent of multi-purpose food shOpping with another dimension of purchase behavior, the extent of food purchase clustering among stores. The findings may be of interest to managements Of supermarkets in shOpping centers who Operate on the general premise that consumers economize on the time and effort required for individual transactions by doing their fOOd and general merchandise shOpping together. The research findings may help supermarket manage- ments to get a better insight into an important aspect of food shOpping behavior of customers--namely, fOOd purchase clustering among stores. Through such a knowledge of cus- tomer behavior, supermarket managements with the choice of apprOpriate marketing devices available to them, might succeed better in achieving a more profitable "customer store loyalty mix" for their stores. Such efforts to improve "customer store loyalty mix" seem imperative for supermarket managements in view of increasing pressures of competition among supermarkets within and without their 'own trading areas' and also in view of the apparent 'similarity' Of supermarkets in shOppers' eyes. 1William L. Garrison §t_§1,, Studies of Highway DevelOpment and Geographic Chapge (Seattle, Washington: University of Washington Press, 1959), Chapter II. 14 Organization The remainder of the dissertation consists Of four chapters: Chapter II reviews the literature relevant to the research problem. The areas which are reviewed include: (1) the early studies in cOnsumer loyalty; (2) some theoret- ical constructs of consumer loyalty; (3) brand loyalty and its relevance for market segmentation research; and (4) empirical research on store loyalty. Chapter III explains the research design and methodology employed in the collec- tiOn and analysis Of the data. The research findings are presented in Chapter IV while Chapter V presents a summary and evaluation of the research hypotheses formulated in Chapter I. In addition Chapter V contains the conclusions of the research and presents some suggested areas for future research. CHAPTER II A REVIEW OF CONSUMER LOYALTY RESEARCH Chapter II presents a review of relevant research in the area Of consumer loyalty. First, a brief presenta- tion Of the early studies which have spurred the interest Of researchers in consumer loyalty phenomena has been made. In the next section, a number Of theoretical constructs in consumer behavior research which have been found useful in explaining the phenomena of consumer loyalty have been presented. The third and fourth sections examine the relevance of consumer loyalty for market segmentation research and present the findings of a number of empirical studies concerned with the important question of identifi- ability Of brand and store loyal customer segments. The final section reviews some studies which emphasized the importance of role perceptions of housewives in influencing fOOd buying decisions and points to their relevance for store loyalty research. 15 16 Consumer Lgyalty--The Early Studies The pioneering work of George Brown1 and Ross Cunningham2 provided the major impetus to much of the later work in the area Of consumer loyalty behavior. Their work focused the attention of marketing researchers to the poten- tial of consumer loyalty as a basis for a profitable market segmentation program for firms. The first major study of brand loyalty was published by Brown in 1952 and 1953. Based on purchase histories Of the Chicago Tribune panel households, Brown examined differ- ences among consumers in terms of brand loyalty for a number of product categories. Brown used the following scheme for the measurement of brand loyalty: Any family making five or more purchases during the year was placed in one of four basic cate- gories, depending upon the purchase pattern shown . . . : 1. Family showing undivided loyalty bought brand A in the following sequence: AAAAAA. 2. Family showing divided loyalty bought brands A and B in the following sequence: ABABAB. 3. Family showing unstable loyalty bought brands A and B in the foIlowing sequence: AAABBB. 1George Brown, "Brand Loyalty--Fact or Fiction?" Advertising Age, Vol. 23 (June 19, 1952), pp. 53-55: (June 30, 1952), pp. 45-47; (July 14, 1952), pp. 54-56: (July 28, 1952), pp. 46-48; (August 11, 1952). PP. 56-58: (September 1, 1952), pp. 80-82; (October 6, 1952), pp. 82-86; (December 1, 1952), pp.176-79: and Vol. 24 (January 26, 1953), pp. 75-76. H 2Ross M. Cunningham, "Brand Loyalty-4What, Where, Homeuch?" Harvard Businegs Review, Vol. 34 (January- February, 1956), pp. 116-128. l7 4. Family showing no loyalty bought brands A, B, C, D E, and F in the following sequence: ABCDEF.1 Using the above classification scheme, Brown observed that a majority of customers concentrate their purchases on a relatively small number of brands and thus exhibit brand loyalty. Brown also noticed that the percent- age Of households that were 'undividedly loyal' varied from 12 percent to 73 percent across products.2 Cunningham3 emphasized the importance of understand- ing consumer loyalty tO manufacturers as well as retailers. His studies broadened the spectrum Of consumer loyalty analysis by focusing upon store loyalty as well as brand loyalty.4 His Operational definition Of brand loyalty was the proportion Of total household purchases represented by the leading single brand used by the household. (An anal- ogous measure was used fOr store loyalty Of households. Among the findings Of Cunningham are: 1. Significant brand loyalty exists within product classes. Loyalty-proneness tendencies across product classes, however, were not significant. 'lGeorge Brown, Op. cit., January 26, 1953, p. 75. 21bid. 3Ross M. Cunningham, "Brand Loyalty-4What, Where, HOMMMuch?" Op. cit. ' 4Ross M. Cunningham, "Customer Loyalty to Store and Brand," Harvard Business Review, Vol. 39 (November-December, 1961). PP. 127-137. 18 2. Families vary widely in their first store loyalty. The store loyalty patterns were reasonably stable over time and not a chance result Of when a partic- ular family happened to be studied. The early studies Of Brown and Cunningham mainly centered around the existence of brand and store loyalties Of consumers. In their studies, consumer purchase data did not support the hypothesis that brands and stores are chosen by consumers on an equiprobable basis, thus pointing to the conclusion that consumer loyalty is a 'real' and reliable phenomena. The same conclusion was arrived at later by Tucker1 who employed an experimental approach to study the formation Of brand loyalty among consumers. In Tucker's experiment, each of a sample of 43 housewives chosen by sociometric methodology was presented four alternative brands on each Of 12 consecutive household deliveries. The loaves were virtually identical except that they were labeled with different 'brand names‘ (L, M, P and H). Based on Tucker's definition, if no brand loyalty were present, it should be expected that 25 percent Of each housewife's purchases will be made fOr each brand. It was found that more than half Of the respondents developed a higher degree of allegiance to one of the four 'brands' than would be 1W. T. Tucker, "The Development of Brand Loyalty," Journalvof Marketing Research, Vol. 1 (August, 1964), pp. 32-35. 19 exPected on an equiprobable basis. The significance Of the experiment lies in the fact that it shows that consumers may become brand loyal even when there is no discernible differ- ence between the branded items other than the brand itself. Consumer Loyalty—-Some Theoretical Constructs The usual purpose Of a theoretical construct is to eXplain some Observed phenomenon. Such constructs evolve from diverse empirical studies and provide a common frame- work for the findings. In turn, they aid in the formulation Of additional hypotheses to be investigated and tested. Much of the empirical research on consumer loyalty, however, has been conducted without the benefit Of a sufficiently developed body of theory in the formulation Of research hypotheses. The choice of research variables has been, for the most part, based on intuitive considerations and explora- tory in nature. The consumer behavior literature Offers some theoretical constructs which appear to hold promise in ex- plaining the phenomena of Consumer loyalty. These constructs focus on some behavioral dimensions in addition to the more usual variables such as price, product quality and store proximity. The theoretical constructs of consumer behavior which appear to hold greatest promise are: i 1. Learning Theory 2. Image Congruence 20 3. Risk Taking Theory 4. Group Influence Each of these is presented along with supportive empirical findings. Learning Theory Learning theory has been advanced by some research- ers as an explanation of brand loyalty behavior of consumers. Four central concepts make up the theory of learning: drive or need, response, cue, and reinforcement. This approach may be summarized briefly in the following terms: . . . Drive impels the subject to respond and the particular response is elicited by a cue. If there were no drive, no response would occur. Thus reSponses are determined by the combination Of drive and cue. If the response is rewarded or reinforced, the response will be repeated when the drive and cue appear together, and thus we can say we have learning. The essence Of learning is this cue-response connection.1 The theory prOposed by Howard and Sheth2 to exPlain consumer brand choice and loyalty behavior has its theoret- ical roots in learning theory. A schematic diagram of the Howard-Sheth paradigm Of brand loyalty is presented in Figure 1. Howard and Sheth focus on the element Of repeat 1John A. Howard, Marketing Theory (Boston: Allyn and Bacon, 1965), pp. 1034104. 2John A. Howard and Jagdish N. Sheth, "A Theory of Buyer Behavior," in Harold Kassarjian and Thomas Robertson (eds.), Perspectives in Consumer BehaviO£_(Glenview, Illi- nois: Scott, Foresman and Company, 1968), pp. 467-487. 21 BRAND CHOICE 0 fl 0 f E “menus or nuvn's "Ann caouce ‘L 2 m " monvss evouo orcmou a BRAND : SET OF summons :' lTY .9. unmannvr g [OVA ' o InAuos 5 O o 5 I S Figure 2-1. Howard-Seth paradigm of consumer brand loyalty. Based on John A. Howard and Jagdish N. Sheth, "A Theory of Buyer Behavior," in Perspectives in Consumer Behavior, ed. by Harold H. Kassarkian and Thomas S. Robertson (Scott, Foresman and Company, 1968), pp. 467—487. 22 purchasing and present a theOry that attempts to portray the dynamics of consumer decision making incorporating concepts of learning theory. The consumer, confronted by repetitive brand choice decisions, simplifies his task by storing rele- vant information and establishing a routine in his decision process. The elements of a buyer's brand choice decision are mentioned as (l) a set of motives, (2) several alterna- tive brand choices and (3) decision mediators by which the motives are matched with the alternatives. The consumer relies on information from his social and commercial envi- ronments and/or his past experience-with similar purchase situations to develop sufficient decision mediators to enable him to choose a brand which seems to have the best potential for satisfying his motives. If the brand proves satisfactory, the potential of that brand to satisfy his motives for subsequent'purchases will be enhanced and the probability of repeat-purchase is increased. With the repeated satisfactory purchases Of a brand, “the buyer is likely to manifest a routine decision process in which the sequential steps in buying are so well structured that an event that triggers the process may also complete it."1 Such a stage in the consumer's purchase process implies high brand loyalty. 1John A. Howard and Jagdish N. Sheth, Op. cit., p. 468. 23 Howard and Sheth believe, however, that a consumer may revert to the search stage from the stage of high brand loyalty. This event is dependent upon the degree of risk perceived by the buyer in the purchase of the brand. In the words Of the authors, "unless a product involves high pur- chase risk, there is a time limit on . . . brand loyalty."1 As in the case Of many frequently purchased products, the consumer may feel bored or become satiated even with a pre- ferred brand and activate his search for new alternative brand choices. The learning theory approach has not been subject tO extensive empirical testing in the marketing context, although such attempts are reported by some researchers2 to be underway. Kuehn's3 probabilistic analysis of Chicago Tribune panel data on household purchases of frozen orange juice showed that repeat brand purchase probabilities in- crease with brand purchase frequency and recency of purchase. The results are consistent with what would be expected on the basis of learning theory. 1John A. Howard and Jagdish N. Sheth, Op. cit., p. 483. 21bid., p. 487. 3.AlfredA. Kuehn, "Consumer Brand Choice as a Learn- ing Process," Journal of Advertiging Research, Vol. 2 (December, 1962), pp. 10-17. 24 Image Congrgence With the growing affluence of the American consumers, marketing researchers have come to realize increasingly that consumer actions are difficult to explain neatly in terms of a 'rational calculus.‘ Price and quality are still impor- tant in the consumer's decision-making process, but the existence and the powerful influence of a host of other intangibles have to be reckOned with at the same time. The significance of a product (or brand) to the consumer Often extends beyond the physical and functional aspects of the product. As Levy Observed, "modern goods are recognized as psychological things as symbolic Of personal attributes and goals, as symbolic of social patterns and strivings." The symbolism associated with a brand (product) in the perception of the consumer is referred to as the brand (product) image and is influenced by a number of factors: socio-cultural influences, group influence, personal charac- teristics of the consumer, person-to-person communications, promotional information and product features. The basic drive Of human beings, in the words of Carl Rogers is "to actualize, maintain and enhance the 2 experiencing organism." In this process Of striving for 1-Sidney J. Levy, "Symbols by Which We Buy," in L. Stockman (ed.), Advancing Marketing Efficiency (Chicago: American Marketing Association, 1958), p. 410. 2Carl R. Rogers, Client-Oriented Therapy (Boston: Houghton and Mifflin Company, 1965), p. 301. 25 self-enhancement, individuals form self-images. .The self- image is "an organized configuration of perceptions of the self which are admissible tO awareness."1 The self-image takes into account one's perceptions about his own qualities and abilities, his relations to his associates and his envi- ronment, and the goals which are desired by himself and which generally enjoy some measure of approval from his 'valued' associates. The image congruence construct of brand loyalty posits that consumers perceive brands as means through which they may achieve their desired self-images and that consum- ers choose and patronize brands whose images (in their per— ception) are most congruent with their self-images. Loyalty to a brand then, persists until the consumers perceive a change either in the brand image or in their self-image. Some empirical studies have been conducted to vali- date the theory that consumer patronage to brands as symbols is patterned in congruent relationships with the consumers' self-image. Birdwelll noticed significant relationships between self-concepts Of buyers and several automobile makes. Grubb, lAlE. Birdwell, "A Study Of the Influence of Image Congruence on Consumer Choice“ (unpublished Ph.D. disserta- tion, University of Texas, 1964). ZEdward L. Grubb, "Consumer Perception of 'Self— Concept' and Its Relation to Brand Choice of Selected Prod- uct Types," in P. D. Bennett (ed.), Marketing and Economic DevelOQment (Chicago: American.Marketing Association, 1965), pp. 419-424. 2 26 in a limited study, found congruence Of self—concept with the brand of beer consumed. Dolich's1 study dealt with two public consumption goods and two private consumption goods and the result appeared to support the theory that consumers tend to relate the brand symbols to self concepts. However, Evans'2 study Of owners of Ford and Chevrolet automobiles failed to discriminate between the owners of the two auto- mobiles in terms Of personality variables. .Evans Observed that "the evidence points neither to strong images attract- ing definite kinds of peOple nor, Specifically to the use of automobiles for satisfying deep inner needs in symbolic terms."3 The image congruence construct has also been found useful in explaining retail store patronage behavior. 'Store image' has been recognized as an important determi- nant Of consumer store loyalty. Store image refers tO "the way in which the store is defined in the shOpper's mind, partly by its functional qualities and partly by an aura of lIra J. Dolich, "Congruence Relationships Between Self Images and Product Brands," Journal Of Marketing Reseprch, Vol. 6 (February, 1969), pp. 80-84. 2Frank B. Evans, "Psychological and Objective Factors in the Prediction Of Brand Choice: .Ford Versus Chevrolet," Journal Of Businesg, VOl. 32 (October, 1959), pp. 340-369. 3Frank B. Evans, "The Brand Image Myth," Business Horizons, Vol. 4 (Fall, 1961), p. 26. 27 PSYchological attributes."l Based on research on shOpping behavior Of consumers in Chicago and its suburbs, Martineau stated that, "the shOpper seeks the store whose image is most congruent with the image She has Of herself."2 Martineau identified social class as an important dimension in the image matching process that underlies con- sumers' retail patronage behavior. ShOppers patronize the stores which reflect the values of the Social class to which they perceive themselves to belong. In a study of 'aggre- gate department Store images,‘ Wyckham3 empirically tested the validity of Martineau's assumption that consumers of different social classes have significantly different per- ceptions Of particular department stores. In the cases of two out Of the three test stores in the study, the findings were supportive of Martineau's assumption, while in the case of the third store there was a commonality of image among all social classes. Wyckham noted, however, that the par- ticular department store had built different types of branch stores that have different images to appeal to different social class groups. Consumers of different social classes might have based their reSponses on their experiences with lPierreMartineau, "The Personality of the Retail Store," Harvard Business;R§yiew, Vol. 36 (January-February, 1958). p. 47. 21bid., p. 48. 3Robert G. Wyckham, "Aggregate Department Store Images: Social and Experimental Factors" (unpublished Ph.D. dissertation, Michigan State University, 1967). 28 the particular branches they patronized and this could have been partially responsible for the Observed commonality of image Of the third store among all social class groups. Risk Taking Theogy Another construct of consumer loyalty focuses on the element of risk taking in consumer decision making. Bauerl considered risk taking as a central concept in explaining consumer purchase behavior. Bauer views consumer actions merely as strategies adOpted by the consumer to deal with the perceived risk in purchase situations: Consumer behavior involves risk in the sense that any action of a consumer will produce conse- quences which he cannot anticipate with anything approximating certainty, and some of which are likely to be unpleasant. . . . Consumers characteristically develOp decision strategies and ways of reducing risk that enable them to act with relative confidence and ease in situations where their information is inadequate and the consequences Of their actions are in some meaningful sense incalculable. Following the reasoning of the risk-taking construct, brand loyalty may be interpreted as a device for reducing the risks in repitive consumer brand choice decisions. Bauer predicted a strong correlation between degree Of perceived risk and brand loyalty if risk is treated as a 1RaymondA. Bauer, "Consumer Behavior as Risk Taking," in Perry Bliss (ed.), Marketing and the Behavioral Sciences (Boston: Allyn and Bacon, Inc., 1963). 21bid., pp. 89-90. 29 combination of uncertainty plus seriousness Of the outcome Of the purchase situation as perceived by the buyer. A similar line of argument underscores the importance of per- ceived risk in influencing the extent of consumer store loyalty. Cunningham1 reported supporting empirical evidence to indicate that repeated purchase Of the same brand is used as a risk-handling strategy by consumers. Arndt2 eXperiment- ing with coffee buyers Observed that high risk perceivers are more likely than those low in perceived risk to be brand loyal and less likely to be interested in adOpting new brands in the same product class. Group Influence Influence Of groups on individual behavior has been the focus Of social psychology and received considerable attention in consumer behavior research. Past research in consumer behavior points to group influence as a significant determinant of brand choice and loyalty behavior of consum- ers. Distinction has been made between two types of groups. 1Scott M. Cunningham, "The Role of Perceived Risk in Product Related Discussions and Brand Purchase Behavior," (unpublished Ph.D. dissertation, Graduate School of Business Administration, Harvard University, 1965). 2Johan Arndt, "Word-OfHMouth Advertising and Perceived Risk," in Harold Kassarjian and Thomas Robertson (eds.), PerSpectivep in Consumer Behavior (Glenview, Illi- nois: Scott, Foresman and Company, 1968), p. 332. 30 The most commonly considered are 'reference groups' which refer to social groups to which a person actually belongs or aSpires to belong or to dissociative groups to which he aSpires not to belong. The other type are 'face-to-face' or informal groups which are characterized by interpersonal interaction over a period of time and a consequent formation Of 'interpersonal bonds Of affect and respect.‘1 Reference groups influence individual consumer behavior in two major ways. Firstly, they influence aspira- tion levels and thus play a part in producing satisfaction or frustration in a purchase situation. Secondly, reference groups influence 'kinds' of behavior by establishing approved patterns of product (brand) acquisition and other aspects Of purchase behavior. Thus they can produce conformity as well as contentment (or discontentment) in a product or brand choice situation. Bourne2 emphasized the importance Of the influence of reference groups in consumer product and brand choice behavior. Consumers patronize products and/or brands which they perceive as 'approved' by their reference groups. 1George C. Homans, Social Behavior: I33 Elementary VForms (New YOrk: Harcourt, Brace and World, Inc., 1961), p. 118. ZFrancis S. Bourne, "Group Influence in Marketing and Public Relations," in Rensis Likert and Samuel Hays, Jr. (eds.), Some Applications Of Behavioral Resgprch (UNESCO, 1957). 31 Bourne, however, recognized that reference groups influence may not be significant in all purchase situations. Patron- age tO particular brands is influenced by reference groups only in the case of what he called 'brand plus' items,1 those for which the brand names are socially conspicuous. The influence Of 'face-to-face' or informal groups on individual members, on the other hand, is effected through the dynamics of interpersonal interaction among members.2 Each member of an informal group has a status and a role within the group. .Informal structuring tends to occur within the group over a period Of time based on the differential status of the members. The more status an individual has within the group, the greater his prestige: the greater one's prestige, the higher he is in the informal hierarchy and the more 'social power' he possesses. Social power has been defined as the total amount of Opinion change one person could induce another to make. The member who has more status and social power than others is generally con- sidered to be the group leader. Small group theory suggests that preferences and loyalty Of informal group members to particular brands of products may be a manifestation Of lIbid., p. 221. 2For a detailed discussion of the concepts Of small group theory, see, George C. Homans, Social Behavior: Ipp. Elementary Forms, Op. cit.: and see also, John A. Howard, Marketing Theory, Op. cit., Chapter V. 32 'intragroup pressure' on members to conform to group norms of behavior. The pressure to conform that a member will 1 Of the eXperience normally increases with the cohesiveness group. Stafford,2 in an eXperimental study, attempted to study how a consumer's brand preferences might be condi- tioned by intragroup communications and the perceptions of brand preferences of fellow members. The results Of the study indicate that the informal group had a definite influ- ence on its members toward conformity behavior with respect to preferred brands.3 Stafford also found that the greater the degree of brand loyalty of the group leader, the higher the percentage Of his group also becoming brand loyal, mOst likely to the same brand preferred by the leader.4 Stafford did not, however, find evidence to support the hypothesis that cohesiveness of a group is a major determinant of the degree of brand loyalty exhibited by the members.5 The theoretical constructs outlined above are some explanations of consumer loyalty phenomena suggested by 1Cohesiveness refers to the attraction a group has for its members. The greater the attractiveness Of the group, the more cohesive the group. 2James E. Stafford, "Effects of Group Influences on Consumer Brand Preferences,“ Jourpgl Of Marketipg,Research, Vol. 3 (February, 1966), pp. 68-75. 3Ibid., p. 75. 4Ibid. 51bid. 33 behavior theories and are important additions to the com- monly known 'rational' factors like price, quality and store proximity. However, it should be noted that probably no Single one of these constructs can completely eXplain con- sumer loyalty behavior and be applicable to all purchase situations. More than one of the above outlined factors may probably underlie loyalty behavior Observed in any specific purchase situation. The particular combination Of the 'critical' determinants which may underlie loyalty phenomena and their relative magnitudes of influence depend upon the Specific purchase situation--the product, the importance of the purchase to the consumer, as well as the personal attributes of the consumer himself. .More research, both at theoretical and empirical levels, is needed tO shed light on the causal influences underlying consumer loyalty phenomena. ngnd Loyalty and Market Seqmentation.Research The strategy of market segmentation has been defined as "the develOpment and pursuit of different marketing pro- grams by the same firm, for essentially the same product, but for different components . . . of the overall market."1 The different component markets are presumably more homoge- neous in relevant consumer characteristics internally than 1Ronald E. Frank, “Market Segmentation Research: Findings and Implications," in Frank Bass, Charles King and Edgar Pessemier (eds.), Applications of the Sciencep in Mpgketing Management (New YOrk: John Wiley and Sons, Inc., 1968). P. 39. 34 the overall market. In a mass-market economy, the strategy of market segmentation helps firms not only to provide prod- uct Offerings that closely match the needs and the tastes of consumers, but also to channel their promotional and other marketing efforts most effectively. A profitable market segmentation program, however, involves a search for mean- ingful bases for segmentation. The pioneering work Of Brown and Cunningham directed the attention Of marketers to con- sumer brand loyalty as a potentially profitable basis for market segmentation policies. Any basis of market segmen- tation has to be evaluated at least against the following criteria:1 1. Identifiability of customer segments: It must be examined whether customers of different segments can be identified in terms of their personal attributes. These personal attributes include characteristics such as socio-economic status, personality and media habits. 2. Differentiability Of purchase characteristics of customer segments: It must be examined whether customers of various segments differ in terms of their purchase characteristics such as average pur- chase level and purchase frequency. 3. Differentiability of promotional elasticities of customer segments: It must be examined whether‘ 1Ibid., p. 43. 35 customers Of various segments differ in their sensitivity to changes in the firm's promotional policies as well as those of the competitors. A number Of empirical research studies designed to evaluate brand loyalty have been conducted with the above criteria in mind. The results of these studies are pre- sented below. Identifiability of Brand Loyal Customer Segments Several investigations have attempted to identify the personal attributes of high and low brand loyal consum- ers for several frequently purchased convenience goods. Cunningham, based on his analysis of purchase data Of a sample of 66 households from a Qpigago Tribupp_panel, reported that socio-economic characteristics had little relation with brand loyalty.1 A study by the Advertising Research Foundation2 dealing with purchase behavior of one-ply and two-ply tissue found virtually no association between personality, Socio- economic variables and household brand loyalty. The total predictive efficacy as measured by the square of the multiple 1ROSS M. Cunningham, "Brand Loyalty--What, Where, How Much?" Op. cit., p. 116. 2Advertising Research Foundation, Age There Consumer Types? (New YOrk: Advertising Research Foundation, 1964). 36 correlation coefficient was 0.05 for one-ply tissue and 0.07 for two-ply tissue. Studies reported by Farleyl focused on the predic- tion of household brand loyalty separately for each of 17 grocery products. The data covered 197 households belonging to the MRCA panel in 1957; the households were made to pre- dict brand loyalty based on knowledge of household income and Size as well as the product consumption rate of each household. The results failed to indicate any significant basis for identifying brand loyal customers. A study conducted by Massy, Frank and Lodahl2 is probably the most extensive investigation Of the association between household brand loyalty and socio-economic and per- sonality attributes. Their analyses encompassed several measures Of brand loyalty and were based on J. Walter Thomp- son's panel data On household purchases Of beer, coffee and tea during 1956-57. The personality data base consisted of scores on the fifteen scales of the Edwards Personal lJohn Farley, "Testing a Theory of Brand Loyalty," Proceggings of the American Mapketinq_A§§0ciation, Wipte£_ Conference, December, 1963, pp. 308-315; and John Farley, "Brand Loyalty and the Economics of Information," Journal Of Business, Vol 37 (October, 1964), pp. 370-381. AWilliam.Massy, Ronald Frank, Thomas Lodahl, Purchage Behavior and Pepponal Attributgp (Philadelphia: University Of Pennsylvania Press, 1968). 37 Preference Schedule (EPPS).l The following results were reported pertaining to brand loyalty:2 1. High incomes and big markets generally mean low loyalty. 2. Husband's endurance score3 is associated with high loyalty for all three products. This is the most stable relationship between personality and brand loyalty behavior. 3. Brand loyalty may have two psychological bases in the wife's personality scores: one based on inde- pendence (autonomy score), and one based on resis— tance and fear Of change (deference and succorance scores). 4. Husband's preferences may also play a strong role in brand behavior in families, considering the number and strengths of the relationships between husband's personality scores and brand behavior. 5. Brand switching behavior may have a psychological basis in needs for affiliation and deference on the 1A. L. Edwards, Manual7for the Edwards Personal Preference Schedule (New YOrk: The Psychological Corpora- tion, 1959). 2WilliamMassy, Ronald Frank, Thomas Lodahl, Pur- chasing,Behavior and Personal Attrlbutes, Op. cit., p. 118. 3In the EPPS, the need for endurance is measured with items such as the following: "to keep at a job until it is finished: to complete any job undertaken; to work hard at a task; to work at a single job before taking on others," etc. Taken together, these items seem to get at a need for completion on the part Of a person. 38 part of the husband, suggesting that husbands in high-switching families are more susceptible to influence attempts. Although the above findings are useful and Signifi- cant in themselves, the results Of the study indicated that only a modest amount of variation in household brand loyalty was explained by persOnal attributes.1 Diffeppntiability Of Purchase Characteripticpiand Elasticities pg Promotion Of Brand Loyal Customers Cunningham2 examined the relationship between aver- age consumption rate and brand loyalty Of households. His analysis indicated that there was little relationship between the two variables. A similar result was Obtained by Massy, Frank and Lodahl.3 One exception is the study report- ed by Kuehn.4 Based on an analysis of frozen orange juice purchases of 650 households from the Chicago Trlbung panel between 1951 and 1953, Kuehn found that brand loyalty lWilliamMMassy, Ronald Frank, Thomas Lodahl, Op. cit., p. 110. 2Ross M. Cunningham, "Brand Loyalty-AWhat, Where, How Much?" Op. cit., p. 116. 3WilliamHMassy, Ronald Frank and Thomas Lodahl, op. cit. 4Alfred Kuehn, "An Analysis Of the Dynamics Of Consumer Behavior and Its Implications for Marketing Manage- ment," (unpublished Ph.D. dissertation, Carnegie Institute of Technology. May, 1968). 39 (measured by repeat purchase probability) was higher for heavy purchasers as Opposed to light purchasers Of the product. Whether brand loyal and nonloyal customer groups differ in terms Of elasticities of promotion was examined by Frank and.Massy.1 If loyalty were successful in building up the resistance Of buyers to switch to other brands in the face of promotional changes in the market, it may be eXpected that the elasticities for loyal buyers would be less than those for nonloyal group. Frank and.Massy's study of the response of a particular brand's market Share in selected markets to changes in pricing, dealing and retail advertising levels revealed no statistically significant differences between the loyal and nonloyal groups in terms of elastic- ities of promotion. The negative character of the results Of the empiri- cal studies reviewed above Show that attempts to establish the relevance Of brand loyalty for market segmentation strat- egy Of firms have not been encouraging SO far. lRonald Frank and William Massy, “Market Segmentation and the Effectiveness Of a Brand's Price and Dealing Poli- cies," Journal of Business, Vol. 38 (April, 1965), pp. 186- 200; and Ronald Frank and William Massy, "Short Term Price and Dealing Effects in Selected Market Segments," Journaliof Marketing Research, Vol. 2 (May, 1965), pp. 171-185. 40 Empirical Research on Store Loyalty Although the managerial need to understand consumer store loyalty patterns was recognized almost a decade ago,1 it has been the subject of limited research only. Cunning- ham2 was the first to broaden the SCOpe of consumer loyalty analysis by focusing on store as Opposed to brand loyalty. Cunningham performed an analysis of store loyalties Of a random sample of fifty families from the Chicago Tribune panel, based on purchases made in seven product categories during 1956. He noticed wide variation in household store loyalty but the store loyalty patterns of individual house- holds were relatively stable Over time. Among Cunningham's other findings3 were: 1. Store loyalty is independent of the total amount spent for food purchases by the family. 2. There is more store loyalty generated toward chain stores than toward specialty stores or independents. lRussell S. Tate, "The Supermarket Battle for Store Loyalty," Journal of Marketing, Vol. 25 (October, 1961), pp. 8-13; and Ross M. Cunningham, "Customer Loyalty to Store and Brand," Op. cit. 2Ibid. 3Ross M. Cunningham, "Customer Loyalty to Store and Brand," Op. cit. 41 3. .Store and brand loyalties are not significantly related.1 4. High store-loyal families are more loyal to the private brands they purchase than are families with low store loyalty. Cunningham's store loyalty analysis was based on household purchase data with respect to a sample Of products rather than an aggregate household food purchases. The mea- sure Of store loyalty employed by Cunningham is the largest prOportion Of food purchases spent in a single store. Such a measure ignores purchases made in the other stores visited by the family. Thus, Cunningham's measure of store loyalty may result in distortions in summarizing a household's pur- chase clustering behavior, especially SO, when the family spreads its food purchases over more than two stores. A new measure of store loyalty2 prOposed in the present study over- comes the above mentioned difficulty. Some studies have attempted to identify important household and personal correlates of store loyalty. In a study of the shOpping behavior of department store customers 1This result was refuted by some later studies. See Tanniru R. RaO, "Purchase Decision Process: Stochastic Models," Journglpof.Marketing Research, Vol. 6 (August, 1969), p. 325; and see also James Carman, ”Correlates Of Brand Loyalty: Some Positive Results," Journal of Marketing Research, Vol. 7 (February, 1970), p. 73. 2See Chapter III, p. 49. 42 in Philadelphia, Blankertz1 found that family income is a Significant correlate Of purchase clustering behavior of customers. Blankertz Observed that "the most important finding of the study was the diSpersion Of trade Of higher income families and relative concentration of trade by low- income families."2 Farley3 factor-analyzed the sample household data used by Cunningham with the addition of some demographic and shOpping activity variables to discern important dimensions Of supermarket choice patterns. He recognized the tendency to Spread purchases over several stores as an important dimension but the "analysis failed to pinpoint characteris- tics of loyal families."4 The only demographic characteris- tics considered by Farley were family Size and income. Massy, Frank and Lodahl's5 study dealt only with pgoduct-specific store loyalty behavior rather than with store loyalty based on aggregate household food purchases. lB.»F. Blankertz, "ShOpping Habits and Income: A Philadelphia Department Store Study," Journal of Marketing, Vol. 14 (January, 1950), pp. 572-578. 2Ibid., p. 574. 3John U. Farley, "Dimensions of Supermarket Choice Patterns," Journal Of Marketing Research, Vol. 5 (May, 1968), pp. 206-208. 41bid., p. 208. 5William Massy, Ronald Frank and Thomas Lodahl, Purchaslpg Behavio; and Pergonal Attributes, Op. cit., p. 120. 43 Their analysis of store loyalty behavior in the case Of three products, beer, coffee, and tea, revealed the follow- ing patterns: 1. Market size and income are associated with low store loyalties. 2. Husband's endurance score is strongly and consistently related to high store loyalty. 3. Husband's deference score and wife's change score are associated with low store loyalty.l The study, however, evidenced a low degree of pre- dictive efficacy of personality and socio-economic variables in explaining prodgct-Specific store loyalty behavior. A study by Enis and Paul2 was aimed at determining whether consumers who exhibit various degrees of store loyalty can be identified by socio-economic and/or psycholog- ical characteristics. The study is one of the few that used total food purchases of households to define store loyalty. The measure of store loyalty is a geometric mean of three commonly employed loyalty indicators: (1) prOportion-Of- budget received by the first choice store, (2) prOportion Of non-switches in first store choice and (3) number of stores in the market not patronized during the survey period. Enis and Paul found that store loyalty tended to be inversely related to educational attainment, and to be higher for blue- collar households than white-collar households. The 1Ibid., p. 110. 2BenM. Enis and Gordon W. Paul, "Store Loyalty: Characteristics Of ShOppers and.Switchers," Southern Journal of Business, Vol. 3 (October, 1968), pp. 267-276. 44 significant personality correlates of store loyalty included consumers' needs for exhibition, achievement, affiliation and deference; and economic and Social values. However the above personality variates accounted for only 13.4 per cent of the total variance in store loyalty. Enis and Paul con- cluded that "for all practical purposes, loyal customers cannot be identified by socio-economic or psychological characteristics."1 The loyalty measure employed by Enis and Paul ignores the purchases made by the shOpper in stores other than the first choice store. Moreover, the measure is not meaningfully defined for shOppers in large metrOpolitan areas since determination of 'the number of stores in the market not patronized' by the shOpper is at best ambiguous. Based on a study of customer loyalty to particular food chains, Carman2 suggested in a recent article that personal characteristics of consumers may be valuable in explaining differences in store loyalty of shoppers. Al- though preoccupation with methodology obscured the precise meaning of some variables employed by him, Carman indicated that perceived roles and interests of housewives are impor- tant predictors of store loyalty. Carman's conclusion was that: lIbid., p. 274. 2James M. Carman, "Correlates of Brand Loyalty: Some Positive Results," Journal of Marketing Research, Vol. (February, 1970), pp. 67-76. 7 45 . . . the most important predictors do pre- sent a profile of the store-loyal and nonloyal shOpper which is meaningful and consistent. The nonloyal consumer is a full-time housewife with a strong interest in cooking and shOpping with the time and means to shOp. The loyal consumer is the busy woman who typically is working to help support a family. She values her time in such a fashion as to devote little attention to entertaining, cooking and being a careful shOp- per. Empirical research on store loyalty appears to indi— cate a lack of consensus as to the usefulness of personal attributes in explaining store loyalty of shOppers. On an .p priori basis, however, it could be reasoned that since store loyalty is a relatively more enduring characteristic of household purchase behavior than brand loyalty (which varies over products), more positive results could be ex- pected in attempts to identify personal correlates of store loyalty. Additional research studies and experimentation with more fruitful dimensions of personal characteristics may be needed before definitive statements can be made about the usefulness of personal attributes in predicting store loyalty of shOppers. 1James M. Carman, Op. cit., p. 70. 46 Other Relevant Research ,A number of marketing scholars1 recognized the housewife's role as the household purchasing agent. A housewife, however, performs a number of other role-related activities like home maintenance, child rearing, entertain- ing in her home, and social and community activities outside her home. The value she attaches to these activities in terms of time and importance may influence her buying behav- ior. Some empirical studies have Shown the importance of role perceptions of housewives on their purchase decisions. Trier2 found that housewives could be distinguished Signif- icantly in terms of their role perceptions with respect to a number of factors influencing food purchasing decisions. Bucklin,3 in a panel study of the food shOpping processes of housewives in Berkeley, identified eight female roles from a factor analysis of some fifty questions on house and job interests and attempted to relate them to the lWroe Alderson,.Marketing Behavior and.Executive Action (Homewood, Illinois: Richard Irwin, Inc., 1957), p. 179; and Henry 0. Whiteside, "Interacting Roles of the Household Purchasing Agent," in Reavis Cox, wroe Alderson and Stanley Shapiro (eds.), Theory in Marketing (Homewood, Illinois: Richard Irwin, Inc., 1964), pp. 270-280. 2Howard Trier, Henry Smith and James Shaffer, "Differences in Food Buying Attitudes of Housewives," Journal of Marketipg, Vol. 5 (July, 1960), pp. 66-69. 3Louis P. Bucklin, "Consumer Search, Role Enactment and Marketing Efficiency," The Journal of Businesp, Vol. 42 (October, 1969), pp. 416-435. 47 food shOpping behavior of housewives. He observed that the most interesting of all the findings of the study was the strategic importance of housewife roles in determining shOp- ping decisions. Bucklin found that the concepts of social position appeared to be less powerful than housewife roles.l There has been very limited research as to the use- fulness of housewife role perceptions in explaining consumer loyalty behavior. The findings of Trier and Bucklin indi- cate the potential fruitfulness of these variables in loyalty research. The present study attempts to examine the useful- ness of role perceptions of housewives in explaining store loyalty patterns. Summary In a mass-market, consumer-oriented economy, firms often embark on a strategy of market segmentation to be able to provide product Offerings that closely match the hetero- geneous needs and tastes of consumers as well as to channel their promotional and other marketing efforts most effec- tively. The pioneering work of Brown and Cunningham directed the attention of manufacturers of frequently purchased con- sumer products to the possibility of employeing consumer loyalty as a profitable basis for market segmentation pro- grams. 1Ibid., p. 435. 48 A vast amount of empirical research has been aimed at examining the feasibility of employing brand loyalty as a basis for market segmentation strategies. The research studies have attempted to identify the personal attributes of brand loyal custOmers and to examine if brand loyal cus- tomers can be distinguished from nonloyal customers in terms of their demand characteristics and elasticities of promo- tion. Findings to date seem to indicate that brand loyal and nonloyal customers are virtually indistinguishable. Thus, research attempts to establish the relevance of brand loyalty for market segmentation strategy have been so far discouraging. Relatively fewer research studies have dealt with store loyalty as compared to brand loyalty. Empirical re- search on store loyalty appears to indicate a lack of con— sensus as to the usefulness of personal attributes in explaining store loyalty patterns of shOppers, although some studies indicated the peor predictive efficacy of socio- economic variables. Recent research suggests the potential fruitfulness of role perceptions of housewives in explaining store loyalty. More research seems to be warranted before definitive statements may be made about the usefulness of personal attributes in explaining store loyalty patterns of shOppers and about the feasibility of its profitable use in market segmentation programs. CHAPTER III RESEARCH DESIGN Chapter III presents the research framework and methodology employed in collecting the data for the research study and testing the research hypotheses generated in Chap- ter I. The first section of the chapter identifies the independent and dependent variables relevant to the research hypotheses. The section introduces a new measure of pur- chase clustering behavior of customers, referred to as 'the entrOpy measure' in the study, which was used as the depen- dent variable in the present research. The second section contains a description of the sampling procedure by which households were selected for the mailing of the question- naires. The third section presents details of the question- naire which served as the research instrument for the study and other details of the data collection process. The final Section gives an account of the statistical analyses rele- vant to the testing of the various hypotheses under investi- gation. 49 50 lgentiflcation of Research Variables lndependent Variables A major part of the research was aimed at investi- gating whether purchase clustering patterns of food shOppers can be identified in terms of selected characteristics of the shOpperS. The characteristics of the shOppers which served as the independent variables for the analysis fall under three general categories. The categories are socio- economic and demographic characteristics of households, self-perceptions of housewives with regard to a number of role-related activities and, food purchasing characteris- tics of the shOpperS. The specific socio-economic and demographic characteristics chosen were: 1. Family income 2. Employment status of the homemaker 3. Educational level of the homemaker 4. Occupational status of the household head 5. Multiple-automobile availability 6. Stage in the family life cycle 7. Family Size 8. Age of the homemaker 9. Number of pre-school age children. The stage in the life cycle variable was a classifi- cation based on the age of the homemaker and the ages of children, if any. Six stages of family life cycle were identified for the study on the basis of whether the family 51 had no children, pre-school age children, or Older children only ppg_whether the homemaker was above or below thirty- five years of age. The second category of independent variables was role-related self-perceptions of the housewives. The role- related activities considered in the study were:1 1. Decorating and cleaning the home 2. Budgeting family finances 3. Rearing and disciplining children 4. Keeping up personal appearance 5. Planning, shOpping and preparing meals 6. Entertaining friends and associates 7. Participating in women's community activities outside the home 8. Planning and arranging recreational activities for the family. The final category of independent variables include some general food purchasing characteristics of homemakers. The specific variables chosen were: 1The role related activity descriptions were adOpted mostly from the role battery used for the Berkeley Food Panel in 1965. .See Louis P. Bucklin and James M. Carman, The Design of Consumer Researcthangls: Conception and Administration of the Berkeley Food Panel (IBER Publications, University of California, Berkeley, 1967), p. 160; the above mentioned role battery was, in turn, closely based on Howard Trier's inventory of role perceptions. See Howard Trier, "Sociological Variables, Personality Traits and Buying Atti- tudes Related to Role Perceptions and Conflicts Among 242 Michigan Housewives" (unpublished Ph.D. dissertation, Michi- gan State University, 1959). 52 1. Total food expenditure of the family 2. Frequency of grocery shOpping 3. weekly in-store shopping time 4. .Extent of multi-purpose food shOpping. In-store shOpping time referred to the amount of time per week that the shopper normally Spends inside food stores buying the food requirements for her family. The extent of multi—purpose food shOpping referred to the fre- x:example, used the prOportion of food purchases made by ‘5 :family in its 'favorite' store as a measure of purchase clustering of the family. As was mentioned earlier, such a me=asure ignores the customer's purchases made in the food Stores other than the favorite store and thus may lead to 53 distortions in summarizing the family's purchase clustering pattern. This is especially so when the family patronizes more than two stores. The research study develOped a summary measure of purchase clustering behavior of shOppers that utilizes in— formation on dollar expenditures made in all the food stores patronized during the study period. .The measure, referred tx> as 'the entrOpy measure of purchase clustering' in the present study, has been adOpted from Shannon's mathematical theory of information. The entrOpy measure of purchase cflnlstering of food shOpperS was defined in the following ternm: Suppose n is the number of food stores in which a faunily made purchases during the study period and pi is the prOportion of the total food purchases that is made in tine ith store. Then, the entropy measure of purchase clus- 'te1:ing of the family (for the study period) is defined by: E = (- 2 pi - Log2 pi) x 100 The measure is non-negative and its value depends on the number of stores the customer patronized for her food heEids as well as on how She Spreads her total food budget C"'€3rdifferent stores. In the simple Situation where the \ "AMMathematical Theory of Communica- 1C..E. Shannon, 27 (1948), pp. 379- tj—On," Bell System Tech_nical Journa_l_, Vol. 423 54 consumer patronized only two food stores, the measure has a near-zero value if the consumer clustered a disprOportion- ately large prOportion of her food purchases in one of the stores and has a maximum value of 100.0 if She spread her food purchases equally among the two stores. A graphical illustration of the values of the entrOpy measure in the above situation is presented in Figure 2. Graphical illus- trations become more complex when the consumer patronizes more than two stores. In general, the more the number of food stores the consumer patronized and the more evenly She Spread her purchases among these stores, the larger will be the value of the entropy measure of purchase clustering (it may be noted that a high value for the entrOpy measure implies low customer loyalty to any single store). The entropy measure of purchase clustering overcomes many of the deficiencies in the existing measures of store loyalty and may be expected to be used more commonly in future consumer loyalty studies. 1The entrOpy measure of purchase clustering was develOped by the writer in his Ph.D. thesis prOposal in February, 1969. At' that time there was no published work which used the entrOpy measure as a quantitative description C>f5 store loyalty patterns of shOppers and the writer was not al‘Nare of any unpublished documents suggesting the entrOpy InSeasure of store loyalty. In two recent papers, however, carman made use of the entrOpy measure in his consumer (Dyalty analysis. See James M. Carman, "Some Insights Into Re asonable Grocery ShOpping Strategies," Joprnag of Marketing, vCD1. 33 (October, 1969), p. 70 (Footnote); and see also J'EsmesM..Carman "Correlates of Brand Loyalty: Some Positive REisults,"g_o_urnal_of Marketing Research, Vol. 7 (February, 1970). p. 75. 55 Maximum = 100 . 0 Entr0py Measure of Food Purchase Clustering 0% 50% 10 0% Percentage of Total Food Purchases Made in Store No. l Fi-sgure 3-1. Graphical illustration of the entrOpy measure of purchase clustering when the shOpper patron- ized only two stores for her food purchases. 56 The entrOpy measure of purchase clustering was used as the dependent variable in the research. In the computa- tion of the entrOpy measure, however, only purchases made in retail grocery stores were considered. Thus, meats purchased in bulk quantities through Special outlets, milk deliveries and milk purchased in dairy stores were not considered in the computation of the measure. Additional Analypis and Relevant Variables For testing hypothesis VI listed in Chapter I, information on additional variables is needed. Information is needed about travel times to the consumer's first and second choice food stores from her home and about the home- maker's perceptions of prices in the two stores. The hypoth- esis suggests that among multiple—store shOppers those who perceive their first and second choice stores as Similar in proximity from home and price image have significantly lower degrees of purchase clustering than others. In the context of the hypothesis, the first and second choice stores were defined as Similar in proximity from home if the homemaker estimates of the driving times to the two stores differ by less than five minutes. The stores were treated as similar in price image if the shOpper places them both in the same position on a semantic differential scale depicting the home- maker's perceptions of store prices. Once these definitions were adOpted, the degrees of purchase clustering of the shOp- pers who perceive their first and second choice food stores 57 as similar in proximity and prices were compared statisti— cally with those of the shOpperS who perceive the stores as dissimilar. Sample Design The Sampling Frame The city of Lansing, Michigan, a community of about 131,500 pOpulation, provided the sampling frame for the research. Besides cost and proximity considerations, the choice was also prompted by the Observation that there has been a noticeable growth in the number of retail food stores in the city over the recent years. The research was primarily concerned with examining relationships between purchase clustering behavior of food shOppers and a number of personal attributes of the shOppers. In the light of this objective, obtaining prOportionate rep- resentativeness of various socio-economic and demographic characteristics in the sample, though desirable, was consid- ered less important than obtaining a random sample of house- holds representing a fairly broad spectrum of socio-economic and demographic characteristics. To meet the time and cost constraints, family income was used as a 'proxy' variable for all household socio-economic and demographic character- istics. The pOpulation was stratified on this variable and quotas of households were randomly sampled from each stratum. The details of the sampling procedure employed in the re- search are presented in the next section. 58 §electlon of Sample Households The sample households in the study were selected according to a multi-stage quota sampling design. .Family income was used as the basis to stratify the pOpulation. Information on median income by census tract for the city of Lansing which was available from 1960 census data,1 was used toward this purpose. Census tracts in the city were grouped into five strata according to median incomes: (1) below $5,000, (2) between $5,000 and $6,000, (3) between $6,000 and $7,000, (4) between $7,000 and $8,000, (5) above $8,000. City blocks in each group of census tracts were enumerated and a random sample of a precalculated number of blocks were selected from each group using random number tables. The precalculated quotas were so determined as to reflect the relative sizes of the strata as well as expected differentials in response rates among different income strata. The groups of randomly selected city blocks provide the sampling frame for the second phase of the sampling pro- cedure. The second phase consisted in the sampling of house- hold units from each of the randomly selected city blocks. Systematic random sample of six households were drawn from each block with the aid of R. L. Polk's glty Dipectopy - Lanpingy Michigan. In the process, care was taken to ‘lU.S. Bureau of the Census, U.S. Census of POpulation and Housing: 1960, Census Tracts, Final Report PHC (l) - 73 (Washington, D.C.: Government Printing Office, 1962). 59 discard any business units drawn as well as residents not comprising family units. Substitutions were made for both such categories. Using the city directory, mailing lists of the sample families who were to receive the question- naires were prepared. The final number of families included in the sample was one thousand. Data Collection Self-administered questionnaires were mailed to one thousand sample households on November 15, 1969. A cover letter that accompanied each questionnaire was addressed per- sonally to the homemaker, and explained the purpose and sig- nificance of the study and urged her COOperation. The cover letters were typed on Michigan State University letterheads. Postage-guaranteed envelOpes with return addresses were enclosed along with the questionnaires, but no monetary or similar incentives were offered to stimulate a high response :rate. The respondents were assured, however, that there was riolway of identifying individuals from the returned question- rlaires. The cover letter is reproduced in Appendix A. The questionnaire develOped for the research was designed to identify significant personal and household correlates of purchase clustering patterns of food ShOpperS. The information sought from the respondents in the question- he ire may be categorized under four broad areas. The first is .iriformation about the homemaker's general food purchasing habits, e.g., her 'normal' frequency of grocery shOpping and 60 extent of her multi-purpose food ShOpping. The second type is information that relates more Specifically to the home- maker's food shOpping experience over the past month. Information was requested about the food stores she visited during the past month and approximate dollar amounts spent in each store. Reliance was placed on the recall of the respondent to provide the estimates of eXpenditures made in each of the patronized food stores. Information was also requested about approximate traveling times to each of the stores from her home. The homemaker was also asked to indicate, on a semantic differential scale, her Opinion about the prices and quality of each food store she patron- ized during the past month. Questions in the third area request the homemaker to indicate, on a semantic differen- tial scale, the time and importance she attaches to a number of role-related activities. The fourth area seeks informa- tion about the socio-econOmic and demographic characteris- tics of the homemaker and her family. The questions per- taining to socio-economic and demographic characteristics were purposefully included as the last in the questionnaire on the assumption that the respondent might lose interest if she were to see routine questions at the beginning of the questionnaire. Special care was taken to limit the size of the questionnaire to four pages. In view of the fact that it was intended for the general public rather than a special- ized audience and that no monetary incentive was involved, it was felt that sending a lengthy questionnaire would 61 definitely mean risking a high non-response rate. The questionnaire is reporduced in Appendix A. The questionnaire was pretested on a very limited scale, mainly for purposes of insuring its general readabil- ity and clarity. The majority of the responses frOm the sample house- holds were received during the first two weeks after the questionnaires were mailed. The number of usable question- naires which were returned during the first two weeks was 238. At the end of the first two weeks, follow up letters were mailed out requesting COOperation from non-respondents. Usable questionnaires which were returned during the third and fourth weeks numbered 87. Analysis of the Data Data‘PgeparaElgp The questionnaires returned by the sample households during the first four weeks provided the data base for the study. The usable questionnaires, which numbered 335 in total out of 1,000 mailed, were coded according to pre- determined classification procedures and the information was transferred to punch cards for computer analysis. The punch cards were verified for accuracy. The computer analysis was primarily confined to the testing of the hypotheses listed in Chapter I. ApprOpriate statistical routines were employed for the purpose. In the process of testing the hypotheses, it was found necessary to 62 generate frequency distributions of a number of household characteristics and to group values of some of these char- acteristics so as to generate cell frequencies large enough to satisfy the assumptions of the relevant statistical tests. Computer Programs for Statlptical Analysis First, an analysis of the sample composition of socio-economic and demographic characteristics was performed using the PERCOUNT computer program1 develOped by the CISSR group at Michigan State University. The program provided a percentage breakdown of the sample households according to each of the socio—economic and demographic characteristics. The composition of the sample with respect to relevant characteristics is tabulated in Appendix C. For testing the Significance of relationships between degree of food purchase clustering and individual personal characteristics (i.e., for hypothesis groups I, II and IV), non-parametric statistical methods were considered to be more apprOpriate than multiple regression analysis. Some personal characteristics are mutually highly correlated and such a situation will lead to the anomaly of multi- collinearity in regression analysis. In the case of multi— collinearity, sampling errors of estimates of regression HMichigan State University, Computer Institute for Social Science Research (CISSR), PERCOUNT, Technical Report NO. 18, May 6, 1968. 63 coefficients may be so large as to make it difficult to draw valid inferences about the statistical significance of individual regression coefficients. For this reason Kruskal-Wallis onedway analysis of variance of ranks was used to test the relationships in hypothesis groups I, II and IV. The nonparametric statistics package newly devel- Oped by the CISSR grOup at Michigan State University pro- vided the computer statistical routinel for the Kruskal- Wallis test. .For testing the significance of relationships between degree of food purchase clustering and groups of variables (i.e., for hypotheses III.B, IV.2 and V), a least squares routine was used. The Michigan State University LS computer program2 on CDC 3600 provided the estimates and tests of Significance of the multiple correlation coeffi- cients correSponding to each of these hypotheses. For hypothesis III.A, a stepwise regression analysis was used to find the relative importance of family income among all socio-economic and demographic variables in eXplaining variations in purchase clustering patterns of food shOppers. AMichigan State University, Computer Institute for Social Science Research, Nonparametplc Chi-quare Tests and Analysis of Vsriance, Technical Report No. 42, June 1, 1966. AMiOhigan State University, Agricultural Experiment Station, Calculations of Least Squares Ppoblemsyon the LS Routine, STAT Series Description No. 7, October, 1968. 64 Michigan State University pspsp computer program1 on CDC 3600 was used for the purpose. For hypothesis IV, Mann-Whitney's U test was used. The nonparametric statistics package of the CISSR group provided the statistical routine.2 AMichigan State University, Agricultural Experiment Station, Stepwise Deletion of Varisbleg; from a_Lreast Sguapss £3;uationp(LSDEL Routine), STAT Series Description No. 9, October, 1968. zMichigan State University, Computer Institute for Social Science Research, Mann-Whltney and Wilcoxon Tests, {Peachnical Report No. 45, September 15, 1967. CHAPTER IV PRESENTATION OF FINDINGS Chapter IV presents the tests of the research hypotheses as they were set forth in Chapter I. Introduc- tory to this, however, a brief examination of the extent of variation in the purchase clustering patterns exhibited by the sample families is made. The first section presents findings relating to the socio-economic and demographic characteristics of food ShOp- pers with varying degrees of purchase clustering. Findings with regard to the predictive efficacy of the socio-economic and demographic variables as a group in explaining varia- tions in food purchase clustering patterns'have also been included in this section. The second section presents find- ings relating to the investigation of other food purchasing characteristics of the respondent families which are hypoth- esized to be significantly related to the extent of food Purchase clustering. Results pertaining to the predictive eff:i.cacy of the selected purchasing characteristics appear in this section. The third section presents findings per- taining to the questions of whether role-related self- perceptions of housewives are uSeful in explaining varia- tlons in the extent of purchase clustering displayed in 65 66 their food shOpping behavior. The final section presents results of the investigation whether similarity of proximity and price perceptions about the food stores patronized by the shOpper all significantly related to the pattern of relative clustering of purchases among these stores. Tabulations supporting the findings have been pre- sented along with each of the hypotheses. In many cases, it was found useful to tabulate the results with values of the entrOpy measure of purchase clustering grouped into four quartiles, since no other natural and meaningful breakdown was apparent. The quartile values for the entrOpy measure of purchase clustering for the sample families are presented in Table 4-1. Statistical significance of a hypothesized relationship is inferred only when the probability of sig- nificance stated in conjunction with the corresponding tabulation is less than 0.05 level. Variationsgl ths Food Purchase Clustering Patterns of the gaggle Families The sample families exhibited substantial variation in their food purchase clustering patterns. Differences in the extent of purchase clustering on the part of the fami- lies as measured by the entrOpy measure and also by the Percentage of total food expenditure spent in the first choice grocery store are depicted by the descriptive sta- tistics presented in Table 4-1. 67 ~.mm "OHHuHmso Home: o.oaa "OAfluumoo Home: H.mh “endows m.mm “Sodom: n.5m "magnumoo Hosea o.mm ”maeuumoo H03OA m.o~ usoHumH>mo pumpcmum H.mm "COHumH>oo pumpsmum ~.¢n acme: m.mh "one: o.ooa ob o.om "modem m.moa 0» o.o “modem MMODm poom .ouauo>mm. msauoumsao may cw ucmmm COHuHomoum mmmsouom mo whommszmmOHucm mmHAHzflh NAAEdm mmB ho mZmNBB HIV mdm<8 68 The entrOpy measure of purchase clustering ranged from 0 to 199.81. A few numerical values of the entrOpy measure corresponding to some Specific shOpping situations may be useful for comparison purposes. It may be noted that when the family patronizes one grocery store exclusively for its food purchases the entrOpy measure is zero. If the fam- ily patronizes two grocery stores during the study period EH16 Spreads its purchases equally among the two, the entrOpy nuaasure assumes a value of 100.0. In the situation where ‘tlae family patronizes four grocery stores during the study period and spreads its purchases equally among these stores, ‘tJnen the entrOpy measures assumes a value of 200.0. These reference values for the entrOpy measure together with the descriptive statistics serve to indicate the substantial \rzariation in the purchase clustering patterns of the sample families. The ensuing sections will examine whether the observed variations in the degree of the purchase clustering may be explained by selected characteristics of the food Sfllcappers. Socio-Economic and Demographic Vaglables The first three groups of hypotheses listed in Chapter I were formulated to identify significant socio- economic and demographic characteristics which can distin- guish between shOppers of varying degrees of food purchase 69 clustering and to examine the overall predictive efficacy of these variables in explaining variations in the clustering patterns. The variables considered are: (1) family income, (2) educational level of the homemaker, (3) employment status of the homemaker, (4) occupational status of the household head, (5) multiple—automobile availability, (6) stage in the family life cycle, (7) family size, (8) age of the homemaker, (9) number of pre-school age children in the family. The results of the tests of these research hypoth- eses will be presented below. F amily lncome Families with higher incomes would presumably have less need for savings that may possibly accrue from multiple- store food shOpping, and thus may be expected to cluster their food purchases to a greater extent than lower income families. The findings concerning the relationship between family income and degree of purchase clustering on the part of food shOpperS are presented in Table 4-2. The data fail to indicate that families in different income groups differ in their food purchase clustering pat- terns. However, for a majority of the families in each of the two lower income strata ($0-—$4,999 and $5,000-$5,999) the degree of purchase clustering ranged in the lower two quartiles indicating a higher store loyalty. On the other hand, among the families with incomes over $15,000 a major- ity of the families (57.8 percent) have the degree of 70 .AO>OH eh.0 um ucmonHcmam AEoooouw mo momummo m an OH.mV m madamzlammeHXm Aeoc Ammo Aeoc Lame Ammo home Amec Aomc Lesa 0.00H o.ooH 0.00H 0.00M o.ooa o.oca o.ooa o.ooa o.ooa Hmuoa xuamaoq 5.0m n.mm n.0m m.cm m.mm h.o~ h.b o.om H.> mawuumoo :uv 0H0uw 30A H.m~ m.om N.>H m.- h.mm h.om H.0v 0.ma h.mm mawuumoo pun v.m~ 0.0m H.0N m.m~ 0.0m 5.0m A.m~ 0.0m o.m~ maauumso pom huamaoq m.ma 0.Hm 0.m~ H.m~ v.Hm 0.0m A.m~ 0.mm o.m~ maeuumoo uma muMum c A: $3 $9 $3 33 $3 $3 $3 33 GS Ho>0 mom.¢aw moo.m~w 000.0Hw 000.0m 000.5» mmm.ow www.mw mmm.¢w meauoumsau unaccusm com On 0» Cu OO 0» Cu Ou Cu 0» Mo ouommmz amonucu 000.maw 000.maw 000.Haw 000.0» 000.0» 000.nw 000.0» 000.mm oEoocH >H«Emm va mqm<8 MMZOUZH NAHSCh wm OZHmMBmDAU mmdemDm m0 mmmOmQ 71 purchase clustering ranging in the upper two quartiles indicating lower store loyalty. In the remaining income strat, the distribution of families in different quartiles of purchase clustering is similar to what may be expected as a matter of chance. The data were not statistically significant. Eggcational Level of the Homemaker Table 4-3 presents the data relevant to the level of education of the homemaker. The educational level of the homemaker was not found to differ significantly for shOpperS of different degrees of purchase clustering. The data indicate, however, that among the homemakers who had 'grade school or less' level of edu- cation, there was a high concentration (41.7 percent) in the first quartile of purchase clustering indicating low store loyalty. However, again, the data were not found to be statistically significant. Epployment Status of the Homemaker Housewives who are employed may have higher Oppor- tunity cost for their time and thus are likely to Show less inclination to do multiple-store food shOpping than non- working housewives. Accordingly, it was hypothesized that there is a significant relationship between the homemaker's employment status and her extent of purchase clustering. The relevant data have been presented in Table 4-4. 72 .Ao>mH om.o um bemoammcoan Asoomonm «0 mmoumdo m on om.e0 m neasmzwamxnauxm :3 $3 Sci find 83 Am: o.oca o.ooa 0.00H 0.00H o.oca o.oca Hmuoa muammoq «.ma 0.mm ¢.m~ m.HN m.mm m.m OHHuHmoo nus ououm 304 m.hm o.m~ m.om 0.mm m.0~ m.mm OHfluHmoO cum m.h~ m.mm m.m~ 0.0m o.m~ 5.0H mawuumoo.ocm muamaoq m.hm m.0m o.om H.0N m.om b.a¢ oaauHmoo uma muoum roam Ase $9 Go so as $8 mmnmwa oumopmuw ommaaoo duodenum Hoonom mmmq Ho mcflumumoao ommnouom poocm>o< HO mmoaaoo meow Hoonom nmam Hoonom mo ouommozqamouucm oumopmno amen meow momuo memEQEOE ecu mo Ho>oq Hmcoflumooom mMNMflSfiZOm mmfi ho AW>MA Hos no.0 on bcnoeeacoen Aaoooonm no awesome e um Hm.~c m headmasamxnsASn Ammo Ros. Abs. Ave. lane Lac Ammo .moec o.ooa o.ooa o.ooA o.ooa o.ooH o.ooA o.ooe o.ooa anode auamaoq H.m~ a.ma n.em o.m~ m.mm o.o m.m~ m.e~ oeauumao rue onoum 30a o.om m.am n.~a e.m~ o.ea A.em v.m~ o.oN oeauumao ohm o.n~ m.~m m.ma o.m~ e.ee m.ea A.e~ o.a~ oaauunao ecu auauaoq n.o~ n.~a m.Am e.m~ v.a~ o.m~ o.ma e.o~ daauumao one ouMum r as so so $8 $3 $3 so as so mucmooum muOXuox mumxuoz muoxuoz mmwcqmsm muoxuoz HoseammmwOHA apocammoMONA mcwuoumoao ommnuuom ooaoaoeocs confident: ooaeaxmuaeom ooaeaxm semen no coupons .deh we «Adamo: seasons powwuom muOuMaquum w mxuwao poem paonomsom on» yo moumum Moccaumaoooo madam qummmaom HEB m0 WDBCBm AflZOHBtADUUO Hm OZHMMBmDAU MmdeMDh ho flflmumfl mlv mqm<8 76 It may be noted, however, that among families headed by unskilled workers, a comparatively low percentage of families fall in the first quartile of purchase clustering indicating high store loyalty. In each of the occupational' categories of semi-skilled workers and prOprietors of small business high percentages of families are represented in the two extreme quartiles of purchase clustering. The first and fourth quartiles accounted for 68.8 and 64.7 percent, respectively, in these occupational categories. These two occupational groups are thus dominated by highly store loyal and highly non-loyal families_with families of intermediate ranges of loyalty under-represented. Mplllpla;§ptomoblle‘Avallability Mobility of the shOpper measured in terms of the number of automobiles available to the family may affect her patterns of food purchase clustering. Multiple-automobile availability contributes to easier access to different sources of food buying and thus would presumably increase the tendency to Spread food purchases among several stores. The findings pertaining to the relationship between a fam- ily's extent of food purchase clustering and the number of automobiles available to it are presented in Table 4-6. The data indicate that the relationship between the extent of food purchase clustering and the number of auto- mobiles available to the family is statistically significant. The table Shows that the percentage of families in the first 77 TABLE 4-6 RELATIONSHIP BETWEEN DEGREE OF PURCHASE CLUSTERING AND MULTIPLE-AUTOMOBILE AVAILABILITYa — L No. of Automobiles Available to the Family Entropy'Measure of Three or Purchase Clustering One Two More (%) (%) (96) High Store lst Quartile 28.1 22.9 5.6 Loyalty 2nd Quartile 22.8 29.9 16.7 3rd Quartile 26.3 21.5 33.3 Low Store 4th Quartile 22.8 25.7 44,4 Loyalty Total 100.0 100.0 100.0 (171) (144) (18) aKruskal-Wallis H (8.27 at 2 degrees of freedom) Significant at 0.016 level. 78 quartile of purchase clustering decreased slightly as we move from one-car families to two-car families. However, a substantial decrease in clustering of food purchases is :noticed when we move to the group of families with three or Inore cars. The percentage of families in the first quartile (of purchase clustering (high store loyalty) was 5.6 percent <:ompared to the corresponding figures of 28.1 and 22.9 for 'the one and two-car families. The percentage of families in the fourth quartile of purchase clustering (low store loyalty) is 44.4 percent compared to the corresponding figures of 22.8 and 25.7 for the other two groups. There were only two families in the sample with no cars and these were omitted from the statistical analysis of the relationship under investigation because of the minimum cell size requirements for the statistical test. Stage of the Familprife Cycla_ The findings relating to the stage in the family life cycle are presented in Table 4-7. The life cycle <:oncept employed was based on the age of the homemaker and 'the ages of the children in the family. The data indicate that a disprOportionately large 19ercentage of families with no children and with the home- !naker under 35 years of age are represented in the first cluartile of purchase clustering imploying a high degree of Eitore loyalty. About 47.6 percent of the families in this Ertage of the family life cycle are represented in the first 79 .Ao>os Aeo.o on osmoamacaan Asooodno mo monsoon m on mm.sac m headsetsmxnsnsm Leosc Ammo Ammo Lame Ammo Lame 0.00M 0.00H 0.00H 0.00H 0.00a 0.00M Hmuoe madmaoa 0.0m v.0m m.mm m.mm m.0m m.m~ oaauumso rue muoum 304 H.m~ o.ov n.0m m.0~ m.H~ m.m~ magnumoo cum m.mm n.- H.NN ~.m~ m.m~ m.¢ oawuumoo USN muammoa H.~N 0.0 0.5m n.0H ~.mm c.5v mafluumso and muoum been $3 Ge 33 $3 $3 63 mm uo>o mm um>o mm uo>o mm Home: mm Hops: mm HopcD mcauoumSHO wmmconsm uomemEOm ummemEOm ummewEOm uwmemEOm ummeQEom uomemEom mo snowmoz,amouucm .SAco soup .smsodaro .sosoasso .sAco coho .sonoaaso .sososaro Iaaco umpao Hoonommum oz IAHBO Hopao Hoonomwum oz msoso omen assess one be women mwAUNU mmHA MAHEflh mm? ZH mudaw Wm OZHmMBmDAU mmflmumbm ho HMMONQ hlfi mammfi 80 quartile. On the other hand, families with no children but with the homemaker over 35 years of age indicated no such distinct patterns of purchase clustering. The distribution of these families in the four quartiles of purchase cluster- ing is close to what might be eXpected under chance. Younger :families (with homemakers under 35) with preschool children (did not exhibit any striking patterns of purchase clustering .although they are slightly over repreSented in the lower two (quartiles (57.7 percent) of purchase clustering. In con- ‘trast, the older families (with homemakers over 35) with pmeschool children indicated a distinct pattern of low <21ustering of food purchases. The percentage of families in this stage of the life cycle that are represented in the first quartile of purchase clustering is zero whereas 77.3 percent of them fall in the upper two quartiles of purchase clustering. Younger families with older children only exhibited a tendency toward low store loyalty as compared to older families with older children only. The distribution of (alder families with older children only among different (quartiles of purchase clustering was not very much different :from what might be expected as a matter of chance. The lrelationship between food purchase clustering and stage in tflne family life cycle was found to be statistically signif- ixzant at 0.041 level. 81 Family-Size The research results pertaining to the size of the family are presented in Table 4—8. The size of the family represents the number of both children and adult members of 'the family. The data indicate that two member families are over :represented in the first quartile of the entrOpy measure of Iourchase clustering (high store loyalty range). About 31.8 19ercent of two member families are represented in the first (quartile. For large size families, the data indicate a tendency toward lower clustering of food purchases among stores. Among six member families, for example, only 13.0 “percent are represented in the first quartile of purchase clustering. The corresponding figures for families with seven or more children is 17.4. .For families of interme- diate sizes, the data did not Show any pattern of purchase clustering Significantly different from what could be expected under pure chance. The data indicated that the (overall relationship between the extent of food purchase <21ustering and family Size was not statistically Significant. .Age of the Homemakep Table 4-9 presents the findings pertaining to the hypothesized relationship between the extent of food pur- Cfllase clustering and the age of the homemaker. 82 deem mien...“ .Ho>oa mo.0 um pomoamacmam Aeoommum «0 mmoumoo m um 0H.mo m madamzwamxmouxm Ammo Ammo Aavo Acme Asmv Aboav 0.00H 0.00M 0.00H o.ooa 0.00H 0.00M Hmuoa muammoa h.am b.am h.am m.m~ m.mH ¢.mm OHHuHmoo.cuw OHODm 30A v.0m v.0m m.ma o.- o.vm «.0N maaunmoo.pnm v.0m m.¢m m.0~ 0.m~ o.am h.ma maaunmso pom muamwoq v.5a 0.ma 0.mm 0.N~ o.sm m.am maauumso uma ououm seem Go Go so so. so . Ge whose o m w m N moaumumoau mmmnouom no n no enammmz.mmouucm MNAHZoa mm.o on nsmoaeesman Asoooonm mo neonate e on mo.ec m neasmzuamxnsnsm Lose Ammo Ammo Rome lame Ammo Aemc Lame o.ooA o.ooA o.ooa o.ooe o.oos o.oos o.ooe o.ooa Hobos madmaoq m.o~ ~.~m ~.sm a.os o.- m.sm m.me m.m~ measumso rue msoum 30a o.m~ H.a~ o.am o.om o.m~ o.o~ o.e~ o.oa espresso ohm ~.o~ o.ms o.om m.e~ m.mm e.m~ e.o~ o.H~ daaonmso can haemaoq o.m~ o.- m.ma o.m~ o.- o.o~ a.o~ m.mm dsaunmso baa duoum roam Axe Axe Ase Lac Ase Axe Axe Axe oo or on me oe mm on mm msasdsnseo masseuse uw>o pom cm can me can 0v 0:0 mm cam on cam mm Coca mo muzmmmz.>mouu:m cmm3uom cmeDOQ cow3uom cow3uom cmmSumm :mw3umm mmmq ummemEOm ecu mo QSOuO mm< mmmxdzwzom HEB m0 mDomO WOC Mm OZHmmBdeU mmdmumbm m0 NmmOmQ mtv mqmf purchase clustering is statistically significant. {Fable 4-12 reveals that as the level of the family's grocery €SXpenditures increased there was lesser clustering of food EDurchases (lower loyalty to any single store). Among fam- idlies with grocery expenditures of $100 or less per month, (alaout 31.2 percent are represented in the first quartile of Fnarchase clustering in comparison with the 22.0 percent in 'tlie fourth quartile. In the case of families who Spent $150 (It more per month on groceries, about 17.1 percent are rep- ressented in the first quartile in contrast with 31.6 percent 1J1 the fourth quartile. Higher food budget thus appears to TABLE 4-12 91 DEGREE OF PURCHASE CLUSTERING BY TOTAL GROCERY EXPENDITUREa Total Grocery Expend it ure/Mon th EntrOpy Measure of $100 or Between $150 or Purchase Clustering Less $100 and $150 More (76) (‘76) (%) I-Iigh Store lst Quartile 31.2 22.0 17.1 Loyalty 2nd Quartile 24.1 27.1 25.0 3rd Quartile 22.7 26.3 26.3 Low Store 4th Quartile 22.0 24.6 31.6 Loyalty Total 100.0 100.0 100.0 (141) (118) (76) aKruskal-Wallis H (6.20 at 2 degrees of freedom) Significant at 0.045 level. 92 be a significant correlate of low clustering Of food pur- chases on the part of the families. Frequency of Grocery Shopping Table 4-13 presents the research results relating to the frequency of grocery shopping by the homemaker. The table shows the distribution of families with different frequencies of weekly grocery Shopping trips among the four quartiles of purchase clustering. The data indicate that the relationship between frequency of grocery ShOpping and food purchase clustering is statistically Significant. The table reveals a distinct pattern indicating that a higher frequency of grocery shOp- ping is associated with lower degrees of purchase clustering (lower store loyalty) and vice versa. Among families who shOp less than once a week for groceries, about 43.9 percent are represented in the first (quartile Of the entrOpy measure of purchase clustering, VVhereas only 12.2 percent of them are represented in the fourth quartile. As the number of weekly grocery ShOpping lzrips increases, it may be noticed that the percent of 1families in the lower quartiles of purchase clustering (higher store loyalty) decreased while it steadily increased Iin the upper quartiles (lower store loyalty). .For example, (anmmg'families who shopped three times a week for groceries <3nly 7.1 percent are represented in the first quartile, VVhile 46.4 percent of them are in the fourth quartile. 93 TABLE 4-13 DEGREE OF PURCHASE CLUSTERING BY FREQUENCY OF GROCERY SHOPPINGa No. of Grocery Shopping Trips/Week Less Four Entropy'Measure of Than or Purchase Clustering One One Two Three More (76) (76) (76) (76) (76) High Store lst Quartile 43.9 28.1 16.0 7.1 14.3 Loyalty 2nd Quartile 17.1 29.8 28.4 7.1 14.3 3rd Quartile 26.8 22.2 23.5 39.3 28.5 JLow Store 4th Quartile 12.2 19.9 32.l 46.4 42.9 Loyalty ' Total 100.0 100.0 100.0 100.0 100.0 (41) (171) (81) (28) (14) aKruskal-Wallis H (26.60 Significant at 0.0000 level. at 4 degrees of freedom) 94 In the case of families who shOpped four or more times a week for groceries, about 14.3 percent are represented in the first quartile in contrast with the 42.9 percent in the fourth quartile. It may be noted that there was only a modest amount of correlation (coefficient of correlation is 0.19) between the frequency of grocery ShOpping and the level of family's food expenditures so that a significant :relationship Of the extent of purchase clustering with one <>f these variables does not automatically imply a signifi- czant relationship with the other. . Extent of Multi-Purpgse Food ShOpping ShOpperS vary in the extent to which they combine ‘their shOpping for general merchandise with their ShOpping :Eor food items. It was assumed that consumers who do not (characteristically mix their general merchandise shOpping Vvith food shopping visit a smaller number of different food astores. Table 4—14 presents the findings relevant to the lIypothesized relationship between the extent of multi- ENarpose food shOpping and food purchase clustering. The data reveal that a substantial percentage of food shOppers do not combine shOpping for food and general '“Merchandise in the same ShOpping with any notable frequency. About 72.2 percent of the respondent families indicated that t11'1ey combine food and general merchandise Shopping 'almost r1ever' or only 'a few times.‘ About 13.7 percent indicated 95 .HO>OH momo.o be ssnosmasoan Asoomosm mo assumes a pm o~.oac m nasamzwsmxnsssm Ammo Lose Lave Loose romeo o.ooa 0.00H 0.00M 0.00H o.ooa Hmuoe muammoq 0.m~ N.~N ¢.m~ 0.wm ¢.ma OHHuHmoo cue muoum 30A v.H~ m.- h.h~ ¢.o~ m.m~ maauumoo cum o.mm m.mm m.m~ o.o” ~.e~ masseuse osm huammoq 0.m~ n.0a ¢.mm m.ma 0.0m mHHuHmSo uma mnoum seem so so so so as m>m3a< maucooomum Mausooomum needs Hm>oz mafiumumsau mmmnonom umOEad auw> , 30m 4 umoaam mo musmmoz.mmouuom means msemmonm Homo: so guacamnonmz Hmumsoo pom moanmoouo nuom How mcammonm mo moamoomuh mOZHmmomm GOOm MmOmmDmIHBADZ m0 BZMHXfi Q2“ OZHmMBmDQU mmémUMDm m0 mmmwmfl ZMNSBWQ mHmmZOHeflflmm fialv WAMQB 96 t:hat they do multi-purpose food shOpping 'very frequently' c>r 'almost always.‘ The data indicate that the relationship between the eextent of multi-purpose food shopping and food purchase <21ustering is statistically significant. The results did riot reveal any distinct trend of decreasing clustering of :food purchases as the extent of multi-purpose food shOpping increased as it was anticipated. The data, however, showed 'that compared to the shOppers who never or almost never did tnulti-purpose food shOpping, each of the other groups exhibited a lower degree of food purchase clustering. In-Store Grocery ShOpping Time The findings relating to the in-store shOpping time are presented in Table 4-15. The data indicate the distri- loution of the respondent families who spend different amounts any single store) and large amounts of in-store grocery shOpping time. Among families who spent less than 60 min- thes per week inside grocery stores shOpping for food about 134.8 percent are represented in the first quartile of pur- <:hase clustering as compared to 14.4 percent in the fourth (quartile. Families who Spent between 60 and 90 minutes in sgrocery shOpping spread their food purchases comparatively ‘tola greater extent. About 30.5 percent of the families in this category are in the fourth quartile of purchase <21ustering as compared to 20.3 in the first quartile. IPamilieS who spent more than two hours per week shOpping :Eor food in grocery stores exhibited a more distinct pattern <>f low clustering of purchases. Only 5.3 percent of these :Eamilies are represented in the first quartile in contrast Vvith the 44.7 percent in the fourth quartile. Almost 80 55ercent of these families are represented in the two upper (quartiles of purchase clustering. Combined Predictive Efficacy of gprchasing,Characteristics Hypothesis IV.2 has been formulated to examine if ‘the selected purchasing characteristics as a group are sig- Iiificant in explaining differences in the purchase cluster- ing patterns of food shOppers. To test the hypothesis, an analysis of variance has been performed testing the 99 s ignificance of the multiple regression function relating ‘tJne entrOpy measure of purchase clustering and the other selected purchasing characteristics of the food shOppers. dIhe associated results are presented in Table 4-16. The ainalysis indicates that the relationship between purchasing (characteristics as a group and the extent of purchase clus- tering is statistically significant. The analysis also :indicated that about 10 percent of the variance in the (degree of purchase clustering has been eXplained by the :Eour selected purchasing characteristics. TABLE 4-16 ANALYSIS OF VARIANCE FOR TESTING THE SIGNIFICANCE OF PURCHASING CHARACTERISTICS As A GROUP3 Degrees Probability Sum of of Mean F of Squares Freedom Square ~Ratio Significance ‘ Ilegression 111251.2 4 27812.81 9.02 < 0.0005 -IError 1017801.3 330 3084.25 UEotal 1129052.6 334 aR2 = 0.10 (R represents the multiple correlation <:oefficient between the entropy measure of purchase clus— ‘tering and purchasing characteristics of families). 100 RoleéRelated Self-Perceptions Recent research in food ShOpping behavior indicated that role perception of housewives have a strong influence on their food purchasing behavior.1 Based on these sugges- tions, it was hypothesized that self-perceptions of house- wives with reSpect to eight selected role-related activities are significantly related to the extent of their food pur- chase clustering. To examine the validity of the hypothesis, an analysis of variance has been performed to test the sig- nificance of the regression function relating the entropy measure of purchase clustering and the eight role—related self-perception scores. The relevant findings are presented in Table 4-17. The analysis indicates that role-related self- perceptions of housewives are not statistically significant as indicators of the extent of their food purchase cluster— ing. The percentage of variance in purchasing clustering that was eXplained by the role-perception scores is only about 3.5. Additional analysis was performed to find if any individual dimension of role-related self-perception of housewives is Significantly related to the extent of food purchase clustering. Kruskal-Wallis analysis of variance has been performed with respect to each role-activity and 1Refer to Chapter II, pp. 46-47. 101 TABLE 4-17 ANALYSIS OF VARIANCE FOR TESTING THE SIGNIFICANCE OF ROLE-RELATED PERCEPTIONS OF HOUSEWIVESa Degrees Probability Sum of of Mean F of Squares Freedom Square Ratio Significance Regression 39811.7 8 4976.5 1.49 0.16 Error 1089240.9 326 3341.2 Total 1129052.6 334 aR2 = 0.04 (R denotes the multiple correlation coefficient between the entropy measure of purchase cluster- ing and the role—related self-perception scores of house- wives). the analysis failed to indicate the significance of apy individual dimension of role perceptions as a correlate of food purchase clustering. Degree of Food Purchase Clustering and PerceptualgSimilaritygof Stopes Consumers who do multiple-store food ShOpping may have different perceptions of the stores they patronize in terms of prices, proximity from home, and quality of merchan- dise. These perceptions are important components of super- market images formed in the minds of food shOpperS. Per- ceptual Similarity of the stores patronized may influence the food purchase clustering behavior of the consumers. Hypothesis VI is formulated to examine if a significant 102 relationship exists between perceptual similarity of stores patronized by consumers and the extent of food purchase clustering among these stores. Only the first and second choice food stores were considered in the analysis, however. It was hypothesized that food shOppers who perceive their first and second choice stores as Similar in prices and proximity from home spread their purchases more equally between the two stores than other shOpperS. Two stores were considered as similar in perceived prices if the homemaker gave them identical ranks on a semantic differential scale with respect to store prices. The stores were considered as Similar in proximity from home if the estimated travel times to the two stores differ by less than five minutes. To test the validity of the hypothesis the entropy measure of purchase clustering has been computed with respect to only the first and second choice stores for each customer who patronized more than one grocery store. .A MannAWhitney one-Sided U test was applied to compare the extent of purchase clustering for the group of shOppers who hold similar price and proximity perceptions about the two stores with that of others. Table 4-18 presents the results of the test. The data indicate that the hypoth- esized relationship is statistically Significant and that shOppers who perceive their two major food stores as similar in prices and proximity do in fact Spread their purchases more equally between the two stores than do other shOppers. 103 mm30q vha nuosuo H0.o m.mmm¢ mmmon mo muesaxoum pom mousse as HmHHEHm mm monoum HOnmE o3u HATED oo>woouom 0:3 mummmozm ooom Cosmoamwcmam umma D monoum 039 Guam mo moguHEZIcch on» on uommmom Suez mamEmm huaaanmnoum mcflumumoau ommnouom mo ouommmz,>mouusm mZOHBQMUmHm NBHSHNOMN QZQ mUHmm NMOBm QZfl OZHMMBmDAO mmflmUMDm m0 Hfimwflfl male mqm<fi 104 Although the above analysis indicated a Significant relationship between similarity of store price and proximity perceptions and the extent of food purchase clustering, similarity of store proximity perceptions alone was not Significant in differentiating between high and low purchase clustering food shOppers. A one-sided Mann4Whitney U test was used to compare the extent of purchase clustering of food shOppers whose two major stores differ or are perceived to differ by less than five minutes in travel time from home with that of other shOppers. The data indicated no statis- tically significant difference between the two groups. Table 4-19 presents the results of the analysis. On the other hand, comparative store price percep— tions by themselves are Significantly related to the extent of the shOpper's food purchase Clustering. A.Mann-Whitney U test indicated that food shOppers who perceived their two major stores as similar in prices spread their food purchases more equally among the two stores than other shOppers. The results are presented in Table 4-20. Additional analysis was performed to examine if comparative perceptions of the quality of meats in the two major grocery stores patronized by the shOpper are signifi- cantly related to the pattern of relative clustering of food purchases between the two stores. Again, an one-sided.Mann- Whitney U test was performed to compare statistically the entrOpy measure of purchase clustering for the shOppers who perceive the quality of meats in the two stores as similar 105 mason on mumnuo 00.0 0.homm surge a: 080: soda muassxoso 5. HmHaEHm mm NUHODN Momma o3» Haonu pm>aooumm 0:3 muommonm poom Cosmoamwcmwm umma D - mUHOum O39 gnaw mo hocuH£31ccmz men on uommmom snag. mHmEmm muaaanmnoum moaumumsao ommnousm mo musmmmz hmouusm I“ Jii mZOHBmflumMm.MBHZHXOMN flmoem 924 OZHMMBmDAU flmH00H0m 0:3 mummmonm 000m Guamowmaomwm Duos D monoum 03B muwm m0 SUCDHDZIGCGE may 0» uommmom rue: mamEmm musasnmnoum maeuoumoao mmmnouom m0 ousmmoszmouucm mZOHBmflumHm MUHMN MMOBm 024 OZHmMBmDAU MmflmUMDm ho Mflmufla ONI¢ MAmGB 107 with that of other shOppers. Table 4-21 presents the results of the analysis. The data indicate that the hypothesized relationship is statistically significant and that shOpperS who perceive their two major food stores as Similar in quality of meats do in fact Spread their purchases more equally between the two stores than other shOppers. Summary Among the nine selected socio-economic and demo- graphic characteristics Of shOppers, only two were found to be significantly related to the extent of their clustering of food purchases among stores. The two significant charac- teristics were: theistaqegln the family llfe cycls.and multiple-automobileiavailability to tha,faglly. The analysis indicated that younger families (homemakers under 35 years of age) with no children tended to cluster their food purchases relatively to a greater extent than older families (homemakers over 35 years of age) with no children. It was also found that older families with preschool age children in the family spread their food purchases among several stores considerably to a greater extent than do younger families with preschool children. However, the data revealed that younger families with older children only tended to Spread their food purchase among several stores to a greater extent than do their counter—part older families. 108 mason mo mumnuo 00.0 m.a¢o¢ mmmon mma homes no Susanna as HmHaEAm mm monoum HOnGE 03D “Hosp om>flmoumm 0:3 muommonm 000m cosmoamwcmwm umoa D monoum 03a Guam «0 mocuw£31osmz mnu 0p DUUQTUM nu“? meEmm auwaanmnoum moanmumoao omenoHom m0 undone: amonucm madfiz mO_MBHQ