Leigh McAlister.

Variety seeking behavior : an interdisciplinary review online

. (page 1 of 2)
Online LibraryLeigh McAlisterVariety seeking behavior : an interdisciplinary review → online text (page 1 of 2)
Font size
QR-code for this ebook


-^^51^%.




VARIETY SEEKING BEHAVIOR:
AN INTERDISCIPLINARY REVIEW*



,Y


5


Dewey
1982 j


LL


^PA


r'.es^^



Leigh McAlister

and
Edgar Pessemier
Sloan W.P. No, 1287-82



March 1982



THE MARKETING CENTER

Massachusetts Institute of Technology

Alfred P. Sloan School of Management

50 Memorial Drive

Cambridge, Massachusetts 02139

(617)253-6615



VARIETY SEEKING BEHAVIOR:
AN INTERDISCIPLINARY REVIEW*



Leigh McAlister

and
Edgar Pessemier
Sloan W.P. No. 1287-82 March 1982



♦Support received from the University of Washington School of Business
Sunimer Research Fund and from the Marketing Science Institute is
gratefully acknowledged.



Leigh McAlister is Visiting Assistant Professor of Management Science,
Sloan School of Management, Massachusetts Institute of Technology,
Cambridge, Massachusetts 02139.

Edgar Pessemier is Professor of Marketing, Krannert Graduate School of
Management, Purdue University, West Lafayette, Indiana 47907.



ABSTRACT
In an attempt to organize the research findings that pertain to the
various phenomena which have been termed "variety seeking" we offer a taxonomy
of varied behavior. Within that framework the literature is reviewed.
Seemingly disparate research traditions; one of which viewed the phenomenon as
inexplicable, the other of which attempted explanation; are shown to be
converging.



- 1 -



VARIETY SEEKING BEHAVIOR:
AND INTERDISCIPLINARY REVIEW

Leigh McAlister and Edgar Pessemier

Variety, the proverbial spice of life, is a much discussed and much

researched topic. Its antecedents, determinants, implications and correlates

have been investigated by psychologists, consumer behaviorists , marketers and

economists. The resulting accumulation of information exhibits seeming

contradictions and inconsistencies. It is our contention that these

discrepancies arise because the term "variety" is used to denote a number of

different phenomena. In this paper we avoid the term altogether as we sort

through the literature that has accumulated under its banner. Toward that end

we offer a taxonomy of causes of varied behavior and review the literature

within that framework.

A TAXONOMY OF VARIED BEHAVIOR

The phenonema of interest in this paper are those mechanisms which lead

individuals to engage in varied behaviors. We include switching among product

variants, switching among service alternatives, switching among various

activities, etc., under the umbrella of "varied behaviors." A taxonomy of

2
explanations for such varied behavior is presented in Figure 1.

- Figure 1 About Here -

The taxonomy suggests that there are two basic schools of thought

concerning varied behavior. One school considers such behavior to be either

inherently inexplicable or, if explicable, to be so complex as to render it

operationally inexplicable. This school focuses research on the probabilities

with which different behaviors will be enacted.

The other school of thought tackles the chore of explanation. We divide

the explanations into two classes: those which view varied behavior as the

result of some other motivation (DERIVED), and those which view variation as a



- 2



motivation in and of Itself (DIRECT). Motivations from which varied behavior
might drive are suggested. It is also suggested that direct varied behavior
might result from internal (INTRAPERSONAL) or external (INTERPERSONAL) causes.

We propose that all these phenomena can cause varied behavior.
Furthermore, any given choice which differs from its predecessor may be the
result of a number of these causes acting in concert. We are not attempting
to invalidate any stream of research. Rather, we attempt to interrelate the
evidence existing in the marketing, consumer behavior, economics and
psychology literatures with respect to varied behavior. Our objective is to
crystalize a few common concepts, identify their antecedents and explore their
implications.

"INEXPLICABLE VARIED BEHAVIOR" AND STOCHASTIC MODELS

Our dichotomization of the literature into schools which consider varied
behavior INEXPLICABLE and EXPLICABLE perhaps overstates their differences.
Those who thought varied behavior too complex for explanation proposed
probabilistic models of choice. The evolution of the probabilistic models has
involved the incorporation of variables to explain the probabilities. So,
little by little, the two traditions are converging.

A major contribution to the tradition of probabilistic prediction of
behavior was Luce's (1959) axiom of choice. His axiom implies a
correspondence between a set of numbers (measures of affect) and a set of
choice objects such that the probability that an object will be chosen is

given by the ratio of the number corresponding to that object and the sum of

3

the numbers corresponding to all of the available alternatives. These

^'
probabilities are often interpreted as long-run frequencies of choice.

In the 1960's, models appeared which focused on different potential

determinants of these "probabilities of choice." Massy, Montgomery, and

Morrison (1970) classified the models as dealing with either: (1) feedback

from past experience with the choice object, (2) the influence of exogenous



- 3 -



market forces, or (3) the impact of factors indigenous to the household
Itself. The first determinant was modelled as the learning brought about
through experience (Keuhn 1962). The second determinant was modelled as a
time trend. The third was captured by assuming, a distribution, across the
population, of the parameters influencing the probability of choice.
(Generally these parameters indicated differences in tastes.)

One stream of subsequent developments in this tradition has switched the
focus to the managerial implications of macro models. Bass and his coauthors
(e.g., Bass 1974; Bass, Jeuland and Wright ; Bass and Pilon 1979) favor
this approach because "the randomness which characterizes individual behavior
tends to be washed out by aggregation" (Bass 1974, p. 9).

Huber and Reibstein (1978) hypothesized that an individual's choice
behavior is explicable and predictable given knowledge of the individual's
ideal points for attributes characterizing the product class. However, since
they posit that those ideal points are subject to random fluctuations, the
final implication is that behavior is not predictable.

Blin and Dodson (1980) cross fertilize multiattribute attitude and
stochastic models. They argue that "the consumer's attitude structure governs
not actual choice on any single purchase occasion but relative frequency of
choice over repeated purchase occasions" (Blin and Dodson 1980, p. 611). The
cross fertilization they propose involves estimating the frequency with which
a behavior will be performed as a function of the importance weights in an
individual's linear compensatory preference function. The data are scaled so
that the importance weights are all positive and so that the weights sum to

one. They estimate the frequency with which a particular behavior will be

4
performed to equal the sum of the importance weights for those attributes

on which that behavior's value is the best available. One implication of this

model is that those behaviors which do not dominate on at least one attribute

will never be performed.



- 4 -



This link between stochastic models and attitude structure is an important
one. It stops short, however, of explaining why and predicting when an
individual will elect to switch behaviors. Givon (1981) proposes a stochastic
model to directly address the question of "why." The sequence of behaviors an
Individual performs, he maintains, is the result of two forces: (1) the
utility derived from the behavior itself and (2) the utility inherent in
switching, regardless of the behavior to which or from which one switches.

The above stochastic models do not address the question of "when" an
Individual will elect to switch behaviors. However, the incorporation of
psychological constructs into the estimation of choice frequencies may
foreshadow the emergence of a third research tradition incorporating the
strenghts of each of the existing traditions.

"EXPLICABLE VARIED BEHAVIOR" AND DETERMINISTIC MODELS

Derived Variation

Models which concern themselves directly with the explanation of varied
behavior can be divided into those which view the variation as being DERIVED
and those which view it as DIRECT. "Derived varied behavior" refers to that
which results from external or internal forces which have nothing to do with a
preference for change in and of itself. We postulate two such forces:
multiple needs and changes in the choice problem.
Multiple Needs

Laurent (1978) developed and validated operational measures of the
construct of varied behavior due to multiple needs. In that research he
decomposed multiple needs into three subcategories: multiple users, multiple
uses, and multiple situations.



- 5 -



Multiple users refers to those cases in which different members of a
household prefer different objects. This heterogeneity of preferences within
the household leads to the selection of multiple objects even if each uses
only a single object.

Multiple situations refers to those cases in which the behavior is
dictated by the situation. When the situation changes, the behavior changes.
"Situations may differ in many ways: the social context of consumption, the
location of consumption, time constraints on consumption, the quantity
consumed, usage convenience (e.g., individual packages of cereal when humidity
is high), variables dependent on emotional reactions, concurrent activities"
(Laurent, 1978, p. 3). Srivastava, Shocker and Day (1978) and Belk (1979)
have extended an approach proposed by Stefflre (1979) to develop a taxonomy of
usage situations. Holbrook (1981) has developed a model incorporating
situation specific ideal points.

Multiple uses refers to those situations in which an object is used in
multiple ways. Examples include the purchase of one type of soup principally
as an ingredient for a casserole and purchase of another type principally for
its use as a course of a meal. One might also purchase one type of baking
soda as a cooking ingredient and a different type as a cleaning agent.
Multiple uses might logically be viewed as a special case of multiple
situations. It is mentioned separately here to maintain consistency with
Laurent (1978).
Changes in the Choice Problem

Varied behavior resulting from changes in the choice problem can be
ascribed to changes in the set of feasible alternatives, changes in tastes or
changes in the constraints facing the individual enacting the behavior.

The feasible set might change for many reasons. New products are being
launched continually and old ones discontinued. A change in the marketing mix
(product, price, promotion or distribution) can be conceptualized as the
addition of a new alternative to the set and the deletion of an old one.

- 6 -



Moving from one neighborhood, region or country to another could change one's
set of feasible choice alternatives. By changing one's perceptions of the
variants, the reduction of uncertainty inherent in learning also changes the
feasible set.

That a change in behavior might result from a change in tastes is
obvious. Tastes could be changed by external influences (e.g., advertising)
or by interanl influences (e.g., maturation).

Finally, a change in the constraints governing choice might lead to the
selection of a different behavior. If one should suddenly have more (or less)
wealth, free time, energy, etc., it would probably affect the selections she
made.

Changes in behavior due to changes in the choice problem are phenomena
with which economists have dealt thoroughly. Two notable economists (Stigler
and Becker 1977, p. 89) claim that "all changes in behavior are explained by
changes in prices and incomes." While many economists would also allow
differences in tastes across individuals and across time, economics research
focuses, typically, on the impact of prices and incomes. The parsimony of
such a paradigm is both a strength and a weakness. It provides valuable
insights into aggregate behavior (Horsky and Sen 1982). However, such models
are often poorly explain and predict a particular individual's choice in a
particular situation. Important work is being done (see Bettman 1979 for an
overview of one such stream of research) to shed light on the internal process
by which changes in the choice problem lead to changes in behavior.

Direct Variation

Psychologists (e.g., Berlyne 1960, Fiske and Maddi 1961, Driver and
Streufert 1964, and Fromkin 1976) suggest that the motivation for varied
behavior may extend beyond multiple needs, and changes in the choice problem.
"Novelty," "unexpectedness," "change" and "complexity" are pursued, it is
proposed, because they are inherently satisfying (Maddi 1968). Furthermore,



- 7 -



one's behavior patterns are typically influenced by those of her peers
(Fromkin 1976). We believe that the inherently satisfying aspects of changing
behavior are caused by forces both internal (INTRAPERSONAL) and external
(INTERPERSONAL) to the individual. The internal forces have to do with the
desire for the unfamiliar, for alternation among the" familiar and for
information. The external forces have to do with needs for group affiliation
and personal identity.
Intrapersonal Motivation

Consumer behavlorists (e.g., Howard and Sheth 1969, Venkatesan 1973) have
linked varied behavior to the existence of an ideal level of stimulation
(novelty, complexity, incongruity, change, etc.). Their arguments are based
on Driver and Streufert's (1964) synthesis of the theories of consistency
(Haider 1946), complexity (Kelly 1955) and optimal arousal (Berlyne 1960).
Driver and Streufert's (1964) theory holds that as stimulation falls below the
ideal level, cognitive action will produce more input (e.g., exploration,
novelty seeking). As stimulation increases beyond the ideal level, cognitive
action will attempt to reduce or simplify input.

Raju's (1980) investigation of the optimum stimulation level suggests
three factors which contribute to the level of stimulation: the desire for
the unfamiliar, the desire for alternation among the familiar and the desire
for information.

The link between the desire for the unfamiliar and the optimal stimulation
level is frequently invoked to explain the purchase of "new products."
Raju (1980) suggests that a favorable attitude toward risk is the primary
motivator of innovativeness.

Alternation among familiar alternatives involves very little risk.
However, it has been pointed out (Venkatesan 1973, Faison 1977) that the level
of stimulation can be raised by switching from one product variant to another,
even if the variant to which one switches is familiar. Early brand loyalty



- 8 -



studies (Tucker 1964, McConnell 1968) and a similar study in social psychology
(Brlckman and D'Amato 1975) document this phenomenon. Subjects in these
studies made repeated choices from a set of initially unfamiliar stimuli. Two
distinct phases of switching behavior were apparent. Initially subjects
systematically explored all stimuli. Later in the experiment subjects tended
to alternate among the elements of their favored subset of the stimuli.
Brlckman and D'Amato (1975) reason that after their initial search, subjects
have v/eak preferences for some stimuli over others. Additional exposure to
more preferred stimuli increases liking of those stimuli. It also creates
boredom. The alternation observed in the later part of the experiments is
attributed to the balancing of these two effects of repeated exposure.

Several models of alternating behavior have been proposed. Givon (1981),
as was mentioned before, suggests that change is rewarding in and of itself
regardless of the object from which or to which one changes.

Farquhar and Rao (1976) propose that reward stems from the configuration
of one's total collection of objects or behaviors. Their "balance" model for
evaluating collections of items allows an item's attributes to have two kinds
of influence on preference for the collection. The first is a simple linear
Increase or decrease (depending on whether the attribute is "desirable" or
"undesirable"). The second has to do with the diversity of values for that
attribute in that collection. If diversity increasess preference for the
collection, the attribute is termed "counterbalancing." If diversity
detracts, the attribute is termed "equibalancing. " The terms in their model
involving "counter-balancing" attributes could be construed as reflecting the
stimulation (novelty, diversity, complexity, etc.) of the collection.
However, they posit a linear relationship between this term and preference for
the collection. Such linearity is not consistent with t^e existence of an
ideal level for stimulation. Rosenfeld (1981) developed a dynamic version of
the balance model for evaluating potential additions to the collection.



- 9 -



Alternation among familiar alternatives might also result from satiation.
Coombs and Avrunin (1977) cites evidence from physiological psychology in
which single peaked preference functions have characterized individuals'
reactions to the constituent attributes of stimuli. A single peaked function,
like that in Figure 2, reflects decreasing, but positive, marginal value for
incremental units of an attribute until the ideal point (or point of
satiation), x-, is reached. The marginal value of additional units of the
attribute becomes negative and continues to decrease as the total store of
that attribute exceeds Xc. Presumably one would change her behavior to
begin acquiring some other attribute as her store of x approached x^,

- Figure 2 About Here -

Similarly, economists intimate that, prices and incomes constant, any
change from one behavior to another is attributable to the decreasing marginal
value of the original behavior (Silberberg 1978, p. 222). Following Lancaster
(1971), the change could be attributed to the decreasing marginal value of the
component attributes resulting from the original behavior.

An important implication of satiation is that the value of a behavior to
an individual must be determined relative to her existing stores or
"inventories" of attributes. Green, Wind and Jain (1972) and Green and
Devita (1974, 1975) developed models of preference for item colJ.ections based
on the premise that there was some sort of interaction among the items in the
collection. McAlister (1979) proposed a model of attribute satiation to
account for that interaction. Her model evaluates the selection of a
collection at a point in time. One would be unlikely to select multiple
replicates of the same item, her model holds. This results from the likely
satiation with attributes in which that item is rich and the relative
attractiveness of other items having attributes only meagerly provided by the
first item.



- 10 -



Jeuland (1978) proposed a model to account for varied behavior over time.
Preference for a behavior at time t, he posits, is a function of not only the
attributes resulting from the behavior but also of one's "experience" with the

Q

behavior. "Variety seeking" (decreasing marginal value) dictates, he
maintains, that preference for a behavior decline as one accumulates
"experience" with it. However, "experience" fades if one goes some period
without enacting that behavior. Therefore, preference for the behavior can
recover. Hagerty (1980) points out that Jeuland' s (1978) model implies that
relative preference among unchosen behaviors do not change over time. Hagerty
demonstrates in an experimental setting that the consumption of a product can
lower preferences for similar products.

McAlister's (1982) dynamic attribute satiation model is similar to
Jeuland' s (1978) except that she postulates that accumulated inventories of
attributes resulting from behaviors rather than accumulated "experience" with
behaviors themselves dictate the selection of different behaviors over time.
This difference speaks to the issue raised by Hagerty (1980). Similar
behaviors definitionally have similar values on relevant attributes.
Therefore enacting one such behavior should have approximatley the same impact

on attribute inventories as enacting a similar behavior. Preference for the

9
similar (but not enacted) behavior will be reduced almost as much as it

would have been had that behavior itself been performed.

The difference between McAlister's (1979) attribute satiation model and

her (McAlister 1982) dynamic attribute satiation model has to do with timing.

The attribute satiation model posits concurrent acquisition/enaction of

behaviors. The dynamic attribute satiation model accomodates behaviors in

successive periods. The dynamic model summarizes a consumption history by the

inventories of attributes that that history generates. In the absence of

consumption, these inventories dwindle over time. Their diminution

corresponds to physiological processing and disposal of such attributes as



- 11 -



sugar content, nutritional value etc., or the "forgetting" of non-
physlologically based attributes (Little and Lodish 1969). At any given point
in time, preferences among behaviors are determined relative to current
inventory levels. Differing inventory configurations at different points in
time combined with the impact of attribute satiation yields an explanation for
alternation among familiar alternatives.

Psychologists and consumer behaviorists have also linked varied behavior
to the acquisition of information . Hirschman (1980) focuses on the search for
new and potentially discrepant information in her study of innovativeness ,
novelty and consumer creativity. As an aside, she hypothesizes that motives
for varying choices among known stimuli do not concern information needs.
Keon (1980) suggests just the opposite. He proposes that "as a consumer
continues to purchase a particular brand, confusion arises as to the true
worth of the brands not purchased." (Keon 1980, pp. 1126-1127; also see
Pessemier 1978). According to Keon's theory, consumers switch brands in order
to refresh their memories about brands not recently purchased.

The study by Raju (1980) suggests that the motive for seeking information
may vary with the level of the simulation ideal. Individuals with high
stimulation ideals may seek information because of a genuine desire to explore
something unfamiliar while individuals with low stimulation Ideals may seek
Information to reduce the risk of trying an unfamiliar product.
Interpersonal Motives

One final motive for varied behavior is the desire for group affiliation
or individual identity. Fromkin (1976) points out that social pressures for
conformity create the need for subtle ways to express individuality. One such
acceptable manifestation of uniquesness is the possession of commodities which
are scarce or unavailable to others. This would account for the varied
behavior stimulated by fashion, whose economic function, according to Robinson
(1961), is to create scarcity. The obvious link between the desire for social



- 12 -



distinctiveness and a proneness to buy "new products" was affirmed in a
dissertation by Szybillo (1973). Fromkin (1976) suggests that innovators are
expressing a socially acceptable manifestation of the desire to see themselves
as different than their peers. We might expect, then, that desire for social
distinctiveness would tend to raise one's ideal level of stimulation.

It should also be noted that varied behavior may also be motivated by the
desire to follow the changing behavior of peers. Here the need for
affiliation will lead to imitation. Most of the literature on social class
and on fashion behavior has emphasized the imporatnce of "keeping up with the
Joneses." The work of Veblen (1899) and Robinson (1961) are illustrative.

THE STIMULATION IDEAL POINT

Because of the centrality of the stimulation ideal to varied behavior we
examine the determinants and implications of that parameter in this section.

A body of literature has developed concerning the determinants of the
ideal point for stimulation. Some of this research (Berlyne 1960) suggests
that there is a hereditary component to that ideal. However, it is clear
(Barron 1953a, 1953b) from adult novelty preference data that learning can
produce differences in these ideal points. Psychological studies have
demonstrated that the intensity and variability of ambient stimulation in
early life will affect preference for intense and varied stimuli later in life


1

Online LibraryLeigh McAlisterVariety seeking behavior : an interdisciplinary review → online text (page 1 of 2)