Professional Documents
Culture Documents
and the number of advertisements read per trip were underlying the phenomena of interest, (2) uncovering
positively associated with wealth and absence of money meaningful alternative research hypotheses, (3) im-
as a child, but negatively associated with the tendency proving the understanding of the empirical results, and
to perceive the benefits of shopping as small and with ( 4) facilitating the selection of appropriate data ana-
the "liberal, worldly, housewife" (Bucklin, 1971). lytic methods. These results are achievable by a model-
Multistore shopping has also been found to be posi- ing process that seeks to define and then evaluate
tively associated with the frequency of shopping, size alternative causal networks.
of the food budget, tendency to mix food and non-food Within the context of modeling store choice behavior
purchases, but negatively associated with that stage in and the role of information search in store choice, the
the life cycle containing families with household heads state of the art is such that, at best, there are only hints
under 35 withovt school-age children (Prasad, 1972). of the types of variables that might be linked together.
Although information search is discussed in the lead- Further, due to the lack of specific theoretical and em-
ing consumer behavior books, the specific nature, cause, pirical attention to store choice behavior, the relative
and impact of store-related search is generally ignored. directions of influence among variables have not been
However, it is possible to gain further insight into the clarified. As a first step towards the specification of a
influences of search behavior by consulting these gen- directional model of store choice behavior, a prelim-
eral summaries. For example, it is suggested that overt inary model was developed to portray the relative di-
search is used: (1) to clarify descriptive and evaluative rections of influences among the variables suggested in
Household/ Buyer
Characteristics ......~---. .
-Location
Attitude ..._ _ _ _...~<
-Demographic
-Role Perceptions of Toward
-Life Style Store Stores
-Per sonality Attributes
In - Store
-Economic
... Information
••
General Opinions I
and Acti vitie s I• _________________ _ _- - - . J
Processing
Concerning
Shopping/Search
tI Strategies for
. ._ _...~ Planning and
L----_________________________ J
Budgeting
I
mational factors hypothesized to influence store atti- collection wave was in week six, or during the second
tudes and store choice, namely price and advertising. week after the new store had opened; the third data
collection wave was in week 19, or 13 weeks after the
THE RESEARCH SETTING new store had opened. In addition to the relative im-
portance attribute scales, subjects rated each store (2
The basis for testing the model in Figure 1 resulted stores in wave 1 and 3 stores in waves 2 and 3) on
from a unique market disruption over time similar to these relative importance attributes using a seven-point
the study by Myers and Nicosia (1970): the market semantic differential scale for each item. Also, in waves
entry of a regional supermarket chain into a duopolistic 1 and 3 respondents completed an activity, interest,
market. Food retailing in the study market had been and opinion (AIO) multiple scale questionnaire. Gen-
dominated by one moderately large independent store eral demographic data were collected dtE'ing wave one,
(Store L), (estimated market share of 25%), and one and additional advertising and coupon dl).ta were col-
regional chain supermarket (Store S), (estimated mar- lected from the panel members in wave three. All data
ket share of 60%). Other than a small number of collection utilized the mail survey techniques and re-
"mom and pop" food stores, these two stores had a spondents were compensated for each survey question-
clear market advantage since the nearest central shop- naire completed. One hundred sixty-nine respondents
ping district (including five chain supermarkets) was completed all three waves.
seven miles away on a congested and obsolete highway.
Since the middle 1960's there had been relatively few
changes in the composition of shopping alternatives Variables
available to buyers with no real change in food shop-
ping alternatives. Thus, prior to August 1973 the mar- Since the primary concern of this paper is with
ket for food retailing could be characterized as stable methodological findings, the following variables are an-
with both retailer and shopper behavioral patterns alyzed and discussed. The particular attribute variables
relatively consistent. were selected because a priori it was expected they
would be the ones most likely to change because of the
relatively heavy promotion by the stores during the
Data Collection Procedures study period. Indeed, the "Advertising helpful" and
Data on subjects' feelings about the relative impor- "Relatively low prices" variables were found to change
tance of attributes of grocery stores using seven-point the most during the study period. (Abbreviations used
rating scales for each attribute were collected from a in Tables 1-4 are shown in parentheses.)
panel of grocery shoppers at three time intervals. The Only the two major, established stores (S and L)
first data collection wave was accomplished four weeks are relevant to the analysis because it is confined to
prior to the opening of the new store; the second data behavioral changes between waves 1 and 3.
22 THE JOURNAL OF CONSUMER RESEARCH
TABLE 1
DIFFERENCES IN CROSS-LAGGED PARTIAL CORRELATION COEFFICIENTS-TOTAL SAMPLE
General Opinions Planning & Budgeting Importance Store Perceptions
Can Com- Spe- Diff Use Adv Low Store L Store L Store S Store S
Save pare cials Store Budget Coupon List Ads Helpful Price AH RLP AH RLP
+* +* + + +* + + -* +
_.
Brand Name
Can Save + + + + -* +* +* +* +* +*
Compare +* + +* + + +* +* + + +* +
+ + +. + +. *
Specials
DiffStore
Budget
+ _.
+ +
+ +
+
+
+
+
-* +. +
+
-* +* +* +* + +
Coupon
List + +
+
+* _._. _. -*
-*
+*
Use Ads
AdvHelpful
ReI Low Price
+* _.
-*
-*
StoreLAH +*
Store L RLP
StoreSAH
TABLE 2
DIFFERENCES IN CROSS-LAGGED PARTIAL CORRELATION COEFFICIENTS-SWITCHER SEGMENT
General Opinions Planning & Budgeting Importance Store Perceptions
Can Com- Spe- Diff Bud- Cou- Use Adv Low Store L Store L Store S Store S
Save pare cials Store get pon List Ads Helpful Price AH RLP AH RLP
Brand Name -* +* + + + + +* -* +
Can Save + + + -* +* + +. + + +.
Compare + +. + + +* + +* +
Specials +* + + +* + +. + +. +
DiffStore + + + + + +* +
Budget + + + + + + +
Coupon + +* +* + + +
List + + + +
Use Ads + +
AdvHelpful +
Rei Low Price
StoreLAH + -*
StoreLRLP +
StoreSAH
+ Indicates variable in left column is the preceding variable (differences in coefficients> 0).
- Indicates variable in top row is the preceding variable (differences in coefficients < 0).
• Indicates difference in correlation coefficients is significant at .IS level.
PATH-ANALYTIC EXPLORATION OF RETAIL PATRONAGE 23
TABLE 3
DIFFERENCES IN CROSS-LAGGED PARTIAL CORRELATION COEFFICIENTS-L-LOYAL SEGMENT
General Opinions Planning & Budgeting Importance Store Perceptions
Can Com- Spe- Diff Use Adv Low Store L Store L Store S Store S
Save pare cials Store Budget Coupon List Ads Helpful Price AH RLP AH RLP
Brand Name +* + + + + -* + + + +
Can Save + + * + + + +
Compare + + +* + +* + + +
Specials + +* +* + + + + + +
Diff Store + + -* + + -* + +
Budget + + + + + +* +
Coupon + +
Ust + + +*
Use Ads +* + +* + + +*
AdvHelpful +* + -* + +
Rei Low Price + + +
Store LAH
Store L RLP
+ +
StoreSAH
TABLE 4
DIFFERENCES IN CROSS-LAGGED PARTIAL CORRELATION COEFFICIENTS-S-LOYAL SEGMENT
General Opinions Planning & Budgeting Importance Store Perceptions
Can Com- Spe- Diff Use Adv Low Store L Store L Store S Store S
Save pare cials Store Budget Coupon List Ads Helpful Price AH RLP AH RLP
Brand Name +* + +* -* + +*
Can Save + + + + + +* + +
Compare +* +* + +* + +* +* + + +* +
Specials + + -* + * + -* *
Diff Store + + +
Budget -* + + + +
Coupon + + + +* +* +*
Ust
Use Ads
+ -* +
Adv Helpful
+ + + +
ReI Low Price
+ + +
Store LAH
Store LRLP
+
StoreSAH
+
+ Indicates variable in left column is the preceding variable (differences in coefficients> 0).
- Indicates variable in top row is the preceding variable (differences in coefficients < 0).
* Indicates difference in correlation coefficients is significant at .15 level.
24 THE JOURNAL OF CONSUMER RESEARCH
tions. The method of path coefficients is primarily use- as long as the variables involved are not highly incon-
ful where most of the variation in a given dependent sistent over time (Pelz and Andrews, 1964).
variable can be associated with the other variables ex-
plicitly included in the system, and where the theory Analytical Complications
underlying the model is well developed (Blalock,
1964). The method of cross-lagged panel correlations The correlational approach of Figure 2 ignores two
is useful when the objective is to isolate probable causal possible confounding effects (O'Brien, 1974). To infer
relations between variables over time. The theory of that A has directional influence over B, it is necessary
the method of path coefficients is succinctly outlined in to account for the effects of the prior level of the de-
Van de Geer (1971) and will not be discussed here. pendant variable B t and the subsequent level of the
predictor variable A t + k on B tH • This can be achieved
by computing the partial correlation coefficients. If
The Nature of the Analysis for Cross-Lagged
l'AtBt+k is sharply reduced by holding AtH constant,
Correlations
then either the serial effect of At, A t + k is strong, or the
Because of the lack of a well-developed theory of simultaneous effect of A Hk , BHk is strong.
store choice behavior and because the data underlying Consequently, in the current study, A will be in-
this study come from a panel, the data were analyzed ferred to exhibit influence over B when l' AtBt+k' AtH:
using the method of cross-lagged correlations. To out- and r.4.t B tH· Bt are statistically significant, and when
I rAtBtH'At+k I - I rBtAt+k'Bt I is statistically signifi-
A and B sequentially influenced each other, there is 1969). However, even in delayed-effect situations much
not, a priori, any obvious cases in which such paired of the cause-effect relationship occurs within a single
effects would move in different directions. Since the wave, leading to an inevitable underestimation of the
objective in this paper is only to assign probable direc- causal relations. Because of this expected underestima-
tion of influence to one variable or the other (or to tion of the causal relations, influence priority of r5 over
neither), then the nature of the relationship (positive r6 (or of r6 over r5) is said to exist if p(r5 = r6) ~ .15.
vs. negative) is not of direct concern. Results of the analysis are presented in the tables
that follow. Tables 1-4 are set up so that the existence
Analytical Procedure of a positive entry indicates causal priority of the vari-
ables listed in the left hand column over the variable
For each of the foregoing variables, means and stan- listed across the top. (Significant I r5 I - I r6 I differ-
dard deviations were calculated for waves one and ences are denoted by asterisks.) Table 5 summarizes
three (the only waves in which all the variables of the results of Tables 1-4 in terms of the four basic types
interest in this paper were measured). Simple correla- of variables studied: Attribute Importance, Store Per-
tion coefficients were calcuated between each pair of ceptions, General Opinions and Activities, and Specific
variables in the manner suggested by Figure 2-six Planning and Budgeting Strategies. Entries to Table 5
correlations per pair of variables. Partial correlation co- indicate the percentage of variable pairs in which vari-
efficients were calculated to control for serial and simul- able types in the left hand column preceded those in
taneous effects as described above.
TABLE 5
DIRECfION OF INFLUENCE BY VARIABLE TYPE
Fraction of Time a Variable of Left Column Types
Precedes Each of the Following
Variable General Planning & Attribute Store
Sample Type Opinions Budgeting Importance Perceptions
Total
General Opinion 15/20 (5/6) 10/10 (SIS) 9/20 (4/8)
Planning & Budgeting 5/20 (1/6) 6/8 (3/3) 5116 (1/4)
Attribute Importance 0/10 (013) 218 (0/5) 018 (0/5)
Store Perceptions 11/20 (418) 11/16 (3/3) 8/8 (SIS)
Switchers
General Opinion 14/20 (3/4) 7110 (3/3) 12/20 (4/5)
Planning & Budgeting 6/20 (1/4) 5/8 (2/2) 10/16 (0/0)
Attribute Importance 3110 (013) 318 (0/2) 0/8 (0/0)
Store Perceptions 8/20 (115) 6/16 (0/0) 818 (0/0)
L-Loyal
General Opinion 11/20 (3/6) 6110 (0/0) 14/20 (1/2)
Planning & Budgeting 9/20 (3/6) 4/8 (1/1) 10/16 (313)
multi-attribute attitude model research issue of whether as much empirical or theoretical attention as other
the beliefs-importance model is superior to the beliefs- variables, yet, in many situations store choice has di-
only model (Bass and Wilkie, 1973). One implication rect influence on resultant brand choice decisions.
of our findings would be that attribute importance is a Buyer information processing is not well-defined in
weak causal agent of store attitudes or store patronage the literature. Although this study has considered spe-
behavior. cific overt search and information utilization variables,
the area of information processing remains an elusive
Store Perceptions-Specific Planning and subject. However, the methodology presented in this·
Budgeting Strategies paper shows promise as a mechanism for. generating
more specific causal hypotheses and . comprehensive
Finding that store perceptions precede specific buyer models of buyer information processing. In turn, better
planning and budgeting strategies supports the notion understanding of patronage behavior and information
that retailer price promotion strategy can have a causal processing should lead to the development of substan-
effect. An apparent effect of the increased price-promo- tive buyer behavior models.
tion activity during the study period was an increase
in the use of coupons, the use of advertisements for de-
veloping shopping lists, and additional incidence of REFERENCES
budgeting activity. It is possible, though, that store
Ring. Consumer Behavior and Application. Boston: Correlation Technique," Psychological Bulletin, 71
Allyn & Bacon, 1974. (January 1969),74-80.
Pelz, Donald C. and Frank M. Andrews. "Detecting Causal Stephenson, P. Ronald. "Determinants of Retail Patron-
Priorities in Panel Study Data," American Sociological age," lournal of Marketing, 33 (July 1969), 57-6l.
Review, 29 (December 1964), 836-48. Turner, Ronald E. "Product Priorities Within a Multiple-
Prasad, V. "Correlates of Multistore Food Shopping Be- Product Marketing Organization," lournal of Market-
havior," lournal of Retailing, 48 (Summer 1972), ing Research, 11 (May 1974), 143-150.
74-81. Van de Geer, John P.lntroduction to Multivariate Analysis
Rich, Stuart U. and Subhash C. Jain. "Social Class and Life for the Social Sciences. San Francisco: W. H. Freeman
Cycle as Predictors of Shopping Behavior," lournal and Co., 1971.
of Marketing Research, 5 (February 1968),41-9. Walters, C. Glenn and Gordon W. Paul. Consumer Be-
Rozelle, Richard M. and Donald T. Campbell. "More havior. Homewood, Illinois: Richard D. Irwin, Inc.
Plausible Rival Hypotheses in the Cross-lagged Panel 1970.