DONALD R.

LEHMANN and YIGANG PAN*

The authors examine how new brand entries affect consumers' considera-
tion sets. A within-subject longitudinal experiment examines several entry
positions into existing markets. The results suggest that new brand en-
tries produce changes in consideration sets toward dominating, compro-
mise, and assimilated brands, away from extreme brands in two-brand
markets, and toward dominating and away from extreme brands in eight-
brand markets. These results are confirmed by a second experiment that
utilizes a between-subject design and markets with six existing brands.

Context Effects, New Brand Entry, and

Consideration Sets

Research on the impact of context effects on choice has suits suggest that new brand entry affects consideration set
been both widespread and informative (Simonson and Tver- membership in a manner similar to their impact on choice;
sky 1992). Similarly considerable interest has been focused specifically, asymmetric dominance and extremeness aver-
on the role of consideration sets in the choice process sion effects are shown to occur. Interestingly, these impacts
(Shocker et al. 1991). In both streams of research, the im- occur for both moderate-size choice sets (six and eight ex-
pact of new products has been an important issue. Our pur- isting brands) and small choice sets (two existing brands).
pose is to merge these research streams by examining the Moreover, being close to an existing brand seems to help a
impact of context effects on consideration sets when new weak brand and hurt a strong one.
products (brands) are introduced.
Certain types of new brand entry have the ability to in- CONSIDERATION SETS
duce violations of some basic assumptions in choice mod- Shocker and colleagues (1991) provide an excellent re-
eling, such as the regularity assumption that an inferior new view of research related to consideration sets. The concept
entry should not increase share of an existing brand (e.g., of choice being limited to a small number of brands was
Huber, Payne, and Puto 1982; Huber and Puto 1983). Nev- formalized by Howard and Sheth (1969, p. 416) as the
ertheless, relatively little is known about the processes that evoked set, which they describe as "those brands the buyer
lead to the findings reported in this area. In developing pos- considers when he/she contemplates purchasing a unit of the
sible explanations for the demonstrated effect, several re- product class." The evoked set recently has been defined as
searchers have looked at the role played by the product at- the set of brands that are evaluated at the point of decision
tributes (Kardes, Herr, and Marlino 1989; Simonson 1989) making (Shocker et al. 1991). The consideration set, on the
or product knowledge (Mishra, Umesh, and Stem 1993), and other hand, is that set of brands being considered at a prior
others have explored the changes in brand perceptions (Pan stage in the choice process and is often portrayed as devel-
and Lehmann 1993) or shifts in attribute weights (Simon- oped by a retrievalprocess from memory (Biehal and Chakrav-
son 1991). We explore the impact of new brand entry on arti 1986; Nedungadi 1990).
consideration sets.
On the basis of previous literature, it seems that brands
In our empirical investigation, we employ several entry
in the consideration set are salient or accessible to consum-
positions in three types of existing markets using both within-
ers. The processing of information on these alternatives is
subject and between-subject experimental designs. Our re-
more active and involved (Nedungadi 1990; Shocker et al.
1991). Moreover, brands in the consideration set are goal
*Donald R. Lehmann is the George E. Warren Professor, Columbia Uni- satisfying in that they satisfy the minimum needs of the con-
versity Graduate School of Business. Yigang Pan is an Assistant Profes-
sor, College of Commerce, DePaul University. The authors acknowledge sumer for the intended use occasion. Alternatives that are
constructive comments from Gregory Carpenter, Avijit Ghosh, Rashi Glazer, apparently not acceptable or not worth considering for the
Bari Harlam, Morris Holbrook, Frank Kardes, David Krantz, Aradhna use occasion would be excluded from further consideration
Krishna, John Lynch, William Moore, Prakash Nedungadi, Thomas Novak, in the choice process. For example, the probability of inclu-
and Robert Pitts on earlier drafts of the article. This research benefits from
funding from the Columbia Business School.
sion of a brand in the consideration set has been modeled
on the basis of the trade-off between consideration cost and

Journal of Marketing Research

364 Vol. XXXI (August 1994), 364-374
Context Effects, New Brand Entry, and Consideration Sets 365

benefit (Hauser and Wernerfelt 1990; Roberts and Lattin HYPOTHESES

1991). We propose that new brand entry affects consideration set
Consideration set research relates closely to research on membership. We expect the dominance (trade-off contrast)
choice process and choice set, which suggests that consum- effect found with respect to choice to occur as well for con-
ers follow a multistage process in which the available sideration set membership. Studies in the attraction effect
brands are reduced to a few for final selection (Bettman area demonstrate that consumers are more likely to choose
1979). This concept was incorporated in several models of a brand when it dominates another brand in the set (e.g.,
consumer choice behavior (e.g., Gensch 1987). This reduc- Huber, Payne, and Puto 1982). Consumers are more likely
tion process is affected by the number of brands presented, to include brands in the consideration set that are clearly
amount of information for each brand, and format of the superior to other brands and are less likely to include
information (Bettman and Park 1980). Alba and Hutchin- brands that are clearly inferior to other brands in the mar-
son (1987) suggest that the size and complexity of consid- ket. Hence:
eration sets is related to consumer expertise. Kardes and col- HI: A dominated entry will increase the likelihood that the (now)
leagues (1993) suggest that the size of the consideration set dominating brand is in the consideration set.
depends on brand heterogeneity. Farquhar and Pratkanis
(1987) suggest that the presence of unavailable alternatives We expect the compromise/extremeness aversion effects
in the choice set affects final choice outcomes. However, it found in previous choice studies to affect consideration mem-
is not clear whether and how a new brand entry affects con- bership as well. It has been found that the addition of a brand
sideration sets. The focus of the current study is to under- to a set of two nondominated brands increases the share of
stand changes in consideration sets induced by the entry of the adjacent brand relative to the share of the nonadjacent
new alternatives in the existing market. brand (Simonson 1989; Simonson and Tversky 1992). A com-
promise alternative appears to be a safer alternative and is
NEW BRAND ENTRY attractive to risk-averse consumers (Huber and Puto 1983).
Furthermore, consumers seem to be able to justify choosing
The topic of new product entry has spawned many re- a compromise alternative better than choosing an extreme
search streams, including modeling the adoption process (Ma- alternative (Simonson 1989). At an early stage in the
hajan, Muller, and Bass 1990), describing appropriate re- choice process, consumers may feel more comfortable con-
sponses to new entry (Hauser and Shugan 1983), and ex- sidering less extreme altematives. Therefore:
plaining first entrant advantages (Carpenter and Nakamoto
1989; Kardes and Kalyanaram 1992). Another research H2: Brands that become extreme after entry will be less likely
stream central to this article concentrates on understanding to be considered; brands that become a compromise after
entry will be more likely to be considered.
how shares realign after entry, with particular emphasis on
unanticipated or irrational behavior. Beginning with Huber, Considerable interest recently has been focused on cate-
Payne, and Puto (1982), many researchers have demonstrated gorization processes (e.g., Cohen and Basu 1987). There are
violations of basic choice axioms such as regularity in that two basic processes that describe how individuals cope with
an existing brand's share increases after the entry of a new new information (Rumelhart and Norman 1972): Assimila-
brand that is inferior. This stream of research is often re- tion occurs when a new concept is integrated into the pre-
ferred to as attraction effect research (Huber and Puto 1983; sent mental schema, and accommodation occurs when a new
Huber, Payne, and Puto 1982; Kardes, Herr, and Marlino mental schema is created or the present schema undergoes
1989; Ratneshwar, Shocker, and Stewart 1987; Simonson substantial modification to interpret the new concept. From
1989, 1991; Simonson and Tversky 1992). Several research- the perspective of the cost of thinking (Shugan 1980) and
ers have proposed possible conditions under which the ef- cost/benefit trade-offs (Hauser and Wernerfelt 1990;
fects take place. For example, the attraction effect can be Roberts and Lattin 1991), brands that are positioned closely
reduced by relatively meaningful alternatives and product to each other would be easier to process and hence receive
class familiarity (Ratneshwar, Shocker, and Stewart 1987). more consideration than the brands positioned less closely
Whereas Huber and Puto (1983) suggest that the effect may to each other. In a choice experiment, a lone altemative is
be limited to cases in which the new entrant is an inferior less likely to be chosen (Glazer, Kahn, and Moore 1991),
brand, Simonson (1989) shows that compromise alternatives possibly because buyers infer desirability from the distribu-
after an extreme entry are more likely to be chosen, which tion of available alternatives. Here, a brand that is positioned
may depend on the need to justify decisions. Summarizing close to another brand is regarded as an assimilated brand,
much of the work on context effects on choice, Simonson and a brand that is positioned far from other brands is a sub-
and Tversky (1992) suggest that two basic tendencies ex- typed brand (Carpenter and Nakamoto 1989; Sujan and
plain most of the results. Specifically, they suggest that in- Bettman 1989). We hypothesize that being positioned close
dividuals prefer alternatives that are clearly better on attrib- to other brands would increase a brand's likelihood of being
utes (trade-off contrast) and are not extreme (extremeness in the consideration set. Thus:
aversion). Our goal is to investigate whether these tenden- H3: A brand positioned close to other brands (assimilated) after
cies also describe the impact of new brand entry on consid- new brand entry is more likely to be in the consideration
eration set membership. set.
366 JOURNAL OF MARKETING RESEARCH, AUGUST 1994

Figure 1 Figure 2
ENTRY POSITIONS IN THE TWO-BRAND MARKETS ENTRY POSITIONS IN THE EIGHT-BRAND MARKETS

Attribute 2 Attribute 2

.A

.1

.2
.4' .3'
.3

.4
.4 .3 .4' .3'
.5
.B
.6
.1 .2
.7

.8
.1' .2 ' .1' .2 '

Attribute 1 Attribute 1

Also, as the number of existing brands increases, most existing brands. Two existing brand markets are included
brands are in compromise positions, so the compromise ef- to match previous research and test Hs.
fect should be weaker in larger markets. The relative size of When the market has only two existing brands, there are
the effects in larger versus smaller existing markets is an eight entry positions for the new entrant: 2 (assimilated and
open issue. On the one hand, one might expect stronger ef- nonassimilated) X 2 (dominating and dominated) X 2 (com-
fects in small markets because manipulations are more ob- promise and extreme). Figure 1 presents the positions of the
vious. On the other hand, in small markets the considera- two existing brands and eight entry positions for the new
tion set might consist of the entire market (e.g., two or three entrant. Brand 1 represents a dominated entrant, brand 2 an
brands) and hence there will be no dominance, compromise, extreme entrant, brand 3 a dominating entrant, and brand 4
or assimilation effects on consideration sets. However, in a compromise entrant. Brands 1,2,3, and 4 are assumed to
established markets in which consumers have predetermined be assimilated and brands 1', 2', 3', and 4' are assumed to
preferences, consumers may quickly reject some alternatives be not assimilated. (This operationalization of assimilation
that fall below a certain cutoff on an attribute (e.g., 40,000 is "objective" rather than perceptual. Hence, the impact of
miles guarantee for car tires). Hence, we propose two more assimilation might be weaker here than it would be if it was
hypotheses: based on subject's perceptions of brand assimilation.)
For the four product categories having eight existing
H4 : The dominance and compromise effects on consideration brands, the eight existing brands are positioned on the effi-
set also will occur in small (two-existing brand) markets. cient frontier in the product space. The new entrant is posi-
H5 : The dominance and compromise effects on consideration tioned in one of the four positions: dominated, extreme, dom-
set will be stronger in smaller markets. inating, and compromise. The assimilation factor was
dropped to simplify the design (Figure 2). Appendix A pro-
STUDY 1
vides the product categories, attributes, and values, based
Subjects on actual brands as rated in Consumer Reports. However,
fictitious names, such as brand S and brand D, were used
One hundred twelve students from a major northeastern for brands.
university participated in the experiment. Most of them at-
tended small sessions with no more than three people in each. Task
Subjects took an average of 40 minutes to complete the task Eight different questionnaires were used in the experiment
and were paid $10 for their participation. to randomize effectively the order of product categories and
Brand Positions
existing brands and the match of entry positions and prod-
uct categories. Each questionnaire contained eight product
Two product attributes were used for all the product cat- categories with two existing brands and two product cate-
egories. Eight product categories have two existing brands gories with eight existing brands.
that are positioned on the efficient frontier in the two- We chose a within-subject design to represent actual con-
attribute space. Four additional product categories have eight ditions for an entry more closely. In a related study, Kardes
context Effects, New Brand Entry, and Consideration Sets 367

and Kalyanaram (1992) investigate order-of-entry effects and able to you?" (Brisoux and Laroche 1981). Both items al-
find that the first entrant advantage may occur because of lowed subjects to check as many or as few brands as they
sequential learning. Interestingly, when subjects are exposed saw fit.
to brands simultaneously, the pioneering advantage tends to Brand choice is measured by asking subjects to pick the
disappear. In our Study 1, there were two time periods. In brand they would buy from all the alternatives (cf. Huber,
the first period, subjects were exposed to ten product cate- Payne, and Puto 1982; Simonson 1989). Brand preferences
gories, each having a set of existing brands. In the second and attribute importance are also measured. Data were col-
period, subjects were exposed to the same ten product cate- lected in the order presented here.
gories and existing brands with a new brand introduced in
each product category. The lag between the two periods was Analysis and Results
a brief interval during which subjects performed an unre- We pooled across product categories to simplify
lated task. presentation.
Results for two existing brands. We first tabulated the per-
Measures
centage of subjects who rated the brand positively on both
Consideration set was operationalized by a two-item meas- consideration set measures (Table 1). The existing brand that
ure. The first item followed the typical measure of consid- is positioned far away from the entrant is labeled brand A,
eration set in previous studies (e.g., Narayana and Markin and the existing brand that is close to the entrant (the target
1975; Nedungadi 1990). Specifically, we asked, "Which brand) is called brand B. In the pre-entry condition, consid-
brand(s) would you consider seriously?" The second item eration set membership averaged 1.30 of the possible 2, in-
of consideration set measure is, "Which brand(s) are accept- dicating that even in a two-brand market not all brands would

Table 1
STUDY 1 RESULTS: TWO-BRAND MARKETS

Consideration set Choice share

Entrant's position Before After Before After
relative to Brand B Brands entry entry entry entry

(3) Dominating and assimilated A 71% 64% 58% 51%

B 54%" 34%" 42%" 0%"
Entrant 73% 49%
Total Size 1.25 1.71
(3') Dominating A 65%b 51%b 46%" 20%"
B 63%" 29%" 54%" 3%"
Entrant 89% 77%
Total Size 1.28 1.69
(l) Dominated and assimilated A 62% 63% 48% 44%
B 68% 73% 52% 55%
Entrant 23% 1%
Total Size 1.30 1.59
(l ') Dominated A 71% 70% 50% 47%
B 63%C 71%C 50% 51%
Entrant 13% 2%
Total Size 1.34 1.54
(4) Compromise and assimilated A 60% 54% 38%c 30%c
B 75%c 66%c 62%" 27%"
Entrant 77% 43%
Total Size 1.35 1.97
(4') Compromise A 62% 58% 46%" 23%"
B 67%b 52%b 54%" 23%"
Entrant 80% 53%
Total Size 1.29 1.90
(2) Extreme and assimilated A 65% 61% 48% 48%
B 68% 65% 52%" 32%"
Entrant 54% 21%
Total Size 1.33 1.80
(2') Extreme A 61% 65% 51% 44%
B 62% 66% 49%c 39%C
Entrant 38% 17%
Total Size 1.23 1.69
Note: N = 112 subjects for each entry condition (aggregated across eight product categories having two existing brands). The number in the brackets refers
to entry positions in Figure 1.
The difference between before entry and after entry conditions is significant at" = p < .01; b = P < .05; c = p <.1.
368 JOURNAL OF MARKETING RESEARCH, AUGUST 1994

be considered by all consumers. After entry, consideration Table 2

sets averaged 1.73 of the possible 3, with the largest gains THE IMPACT OF NEW ENTRY ON CONSIDERATION SETS:
when compromise entries were introduced and the smallest REGRESSION ANALYSIS (STUDY 1)
gains when dominated entries were introduced.
When the entrant dominated B, significantly fewer sub- Independent Two existing Eight existing
jects would include B in the consideration set both in the variables brand markets brand markets
case of assimilated entry (35% versus 54%, p < .01) and Intercept .96 .96 .88
nonassimilated entry (29% versus 63%, p < .01). When B (14.47)* (14.50)* (11.00)*
dominated the entrant, more subjects would include B in Assimilated .04
(Z.33)t
the consideration set, but the difference was significant only Compromise .05 .09 -.05
in the case of the nonassimilated entry (71% versus 63%, (2.37)t (3.05)* (-1.69)t
p < .1). Thus, HI is partially supported. When the entrant Extreme -.14 -.20 -.13
was in a compromise position, fewer subjects would con- (--6.28)* (-7.05)* (-4.45)*
Dominating .09 .14 .06
sider B both in the case of assimilated entry (66% versus (3.96)* (4.96)* (2.16)t
75%, p < .1) and nonassimilated entry (52% versus 67%, Dominated -.41 -.44 -.22
p < .05), supporting Hz. Interestingly, when B becomes a (-18.17)* (-15.64)* (-7.73)*
compromise brand, its likelihood of being in the considera- Compromise x assimilated -.02
(-.50)
tion set does not increase. Here again we see an asymme-
Extreme x assimilated .15
try: Being in a disadvantaged position hurts consideration (4.05)*
but being in a favored position does not increase it after Dominating x assimilated -.05
entry. Finally, the entrant was more likely to be in the con- (-1.40)
sideration set when it was in a dominating or compromise Dominated x assimilated .10
(2.52)t
position and less likely when it was in an extreme or domi- ModelR2 .18 .19 .16
nated position. Brand A's consideration set membership, as
expected, was less affected by the new entries than brand Note: Values in parentheses are t-statistics. * = p < .01; t = p < .05; t =
p < .1. Data are aggregated across eight product categories having two
B's. existing brands.
The impact of new brand entries on consideration sets is
further assessed through regression. The dependent variable other factors. The four interaction terms (between assimi-
is consideration set membership. The two measures of con- lated and dominating, dominated, compromise, and extreme)
sideration set membership are correlated (r = .75 for cate- explain significantly more than just the main effect alone
gories having two existing brands and r =.81 for categories (F[3,4465] = 6.08, p < .01). The interactions between as-
having eight existing brands), so a single index for consid- similated and extreme and between assimilated and domi-
eration set was created by averaging the two items. The in- nated are positive and significant (p < .01 and p < .05). This
dependent variables are dummy variables representing each suggests that for a brand in an undesirable position (extreme
brand's position relative to other brands defined on five di- or dominated), being close to a brand in a more desirable
mensions: assimilated, compromise, extreme, dominating, position (dominating or compromise) increases its likelihood
and dominated. The before-entry brands and the nontarget of being in the consideration set. Specifically, being close
existing brand were coded 0 across all the dimensions. After (assimilated) cancels out 75% of the negative effect of being
new brand entry, the entry and target brands are coded 0 or extreme but less than 25% of the effect of being dominated.
1 on the five dimensions, depending on their relative posi- As expected, for a brand in a desirable position (dominat-
tions. For example, if the entrant dominates the target exist- ing or compromise), the impact of assimilation is less dra-
ing brand, the entrant is coded 1 on the dominating dimen- matic (the two interaction terms are not significant). How-
sion and 0 on other dimensions, and the target brand is coded ever, the signs for being assimilated to an undesirable brand
1 on the dominated dimension and 0 on other dimensions. are negative, suggesting that the gains from being assimi-
The data are pooled across subjects, product categories, lated might be more than offset by losing distinctive
and before/after entry conditions, with each combination of superiority.
subject, product category, and before/after entry conditions The choice shares of brand A, brand B, and entrant also
treated as a single data point (We generated 4475 usable are tabulated in Table 1. As expected, brand B' s shares were
observations: 112 subjects by eight categories by two before- affected in the predicted directions and were significant in
entry brands plus three after-entry brands minus cases with six of the eight entry conditions. Not surprisingly, the
missing data.) To control for the partially repeated measure choice results are similar to those involving consideration
nature of the design, both subject and product dummies are set membership.
included in the model as well. Changes in existing brands' consideration set member-
The compromise, extreme, dominating, dominated, and ship. The results reported in Table 2 come from two sources:
assimilated coefficients are significant and in the directions the entrant (second period only) and the existing brands be-
hypothesized (Table 2), again supporting HI' Hz, and H3 . fore and after entry. An analysis that focused only on the
The impact of assimilation may vary on the basis of whether reactions to different new entries shows, unsurprisingly, that
the assimilated brand is in a dominated, dominating, com- the entrant does well when it is in a dominating or compro-
promise, or extreme position. We replaced the assimilation mise position and poorly when it is in an extreme or domi-
main effect with interactions between assimilation and the nated position.
Context Effects, New Brand Entry, and Consideration Sets 369

Table 3 consideration (p < .1). A dominated position has a negative

THE IMPACT OF NEW ENTRY ON CHANGES IN effect on consideration (p < .01). An extreme position has a
CONSIDERATION SETS negative impact on consideration (p < .01). However, here
a compromise entry has a marginally significant negative
Independent Two existing Eight existing impact (p < .1). In this case, the eight existing brands are
variables brand markets brand markets spread out on the efficient frontier, spanning the meaning-
Intercept -.03 -.03 .02 ful range on product attributes. Hence, it is much harder for
(-.33) (-.33) (.27) a compromise brand to be distinct than for a brand in a mar-
Assimilated .05 ket of two existing brands.
(1.75H Focusing on the impact of entry on changes in the con-
Compromise .03 .08 -.01
(.98) (2.13)t (-.45) sideration sets for existing brands, we see the results are sim-
Extreme -.08 -.08 -.04 ilar to those of the two-brand case, though somewhat weaker.
(-2.37)t (2.1O)t (-1.60)§ Specifically, the dominating effect, though still positive, fails
Dominating .08 .09 .02 to achieve significance in the eight-brand market.
(2.37)t (236)t (.93)
Dominated -.24 -.30 -.09
By comparing the results of the two- and eight-brand mar-
(-7.46)* (-7.69)* (-3.32)* kets, we see support for both H4 and H5. Clearly the domi-
Compromise x Assimilated -.06 nating, compromise, and assimilation effects occur for the
(-1.12) two-brand markets. With the exception of the compromise
Extreme x Assimilated .06 effect, the directions of the effects are the same in the eight-
(1.11)
Dominating x Assimilated .01 brand markets. Moreover, the magnitude of the effects are
(.26) slightly greater for the extreme and dominating effects and
Dominated x Assimilated .17 noticeably larger for the dominated effect in the two-brand
(3.25)* markets.
ModelR2 .14 .15 .11
Note: Values in parentheses are t-statistics. * = p < .01; t = p < .05; t = STUDY 2
p<.l;§=p<.ll.
Data included only the existing brands. The dependent variable is (After Motivation
Entry Consideration) - (Before Entry Consideration). There are several limitations in Study 1. First, the within-
subject design might produce a demand effect on subjects,
The within-subject design allows for an analysis of the
because the first period's response tends to influence later
changes of existing brands over the two periods. The de- responses. Second, contamination of consideration set,
pendent variable is the within-subject after-entry considera- choice, and preference measures is a possibility. Third, all
tion set membership minus the before-entry one. For cate- the products in Study I had two attributes. It is apparent
gories having two existing brands (Table 3), the results show that for more complex products, consumers are likely to use
that after the entry, an existing brand in an assimilated posi- more than two attributes in their consideration. In Study 2,
tion (p < .1) or a dominating position (p < .05) is more likely we attempt to address these issues.
to be in the consideration set. An existing brand in an ex-
treme position (p < .05) or a dominated position (p < .01) is Design
less likely to be in the consideration set. The compromise A pretest on 17 subjects was carried out on eight prod-
effect is positive but not significant. Again, the interactions ucts having four to six attributes taken from Consumer Re-
between assimilation and the other factors contribute sig- ports.i In the pretest, product involvement, attribute famili-
nificantly more to model predictability than to the main ef- arity, attribute meaningfulness, attribute importance, and at-
fect of assimilation alone (F[3,1780] = 3.23, p < .05). The tribute specification familiarity were measured.
interaction between dominated and assimilation is positive On the basis of the pretest results, we chose two high-
and significant (p < .01). Thus, entry effects influence involvement (personal computers and compact sedans), two
changes in consideration set membership. moderate-involvement (air conditioners and orange juice),
Results for eight existing brands. The average considera- and two low-involvement (light bulbs and stereo speakers)
tion set size for the eight-brand market is 4.19 before new products. Compact sedans, orange juices, and light bulbs
entry. After the entry of the ninth brand, the size is 4.36 for were described on two attributes, and personal computers,
a dominating entry, 4.18 for an extreme entry, 4.08 for a air conditioners, and stereo speakers were described on three
dominated entry, and 3.98 for a compromise entry. Note that attributes. The product categories, attributes, and brand spec-
a compromise entry results in a decrease in consideration ifications used in Study 2 are familiar and important to most
set size. On average, .84 fewer existing brands were consid-
ered after a compromise entry. The percentage of respon- IThe following eight products and attributes were pretested: Compact
sedans: reliability, fuel economy, price, ride comfort, and style; light bulbs:
dents that would consider the new entry is 86% for a domi- light output, usage life, price, cost per 1000 hours, and efficiency; air con-
nating entry, 63% for a compromise entry, 30% for an ex- ditioners: cooling capacity, energy efficiency, price, moisture removal, and
treme entry, and 7% for a dominated entry. noise; stereo speakers: accuracy, power, price, bass capacity, and style; walk-
ing shoes: comfort, support, price, weight, and style; orange juice: taste,
Regression analysis produces results that are generally di- nutrition, price, and type; personal computers: base memory, speed, price,
rectionally similar to those from the two brand markets hard disk capacity, and warranty; and laundry detergents: stain removal,
(Table 2). A dominating position has a positive impact on brightening, price, softening, and type.
370 JOURNAL OF MARKETING RESEARCH, AUGUST 1994

Figure 3 ditions (dominating, dominated, extreme, and compromise).

ENTRYPOSITIONS INTHETWO-ATTRIBUTE, SIX-BRAND MARKETS Two hundred twenty graduate business students participated
in Study 2 as part of a class requirement.
In spite of efforts to make both ends of the efficient fron-
Attribute 2
tier equally desirable, in each of the markets a slight prefer-
ence was evident for one end (e.g., high on attribute I and
low on attribute 2 was preferred to low on attribute 1 and
high on attribute 2). Although this may reduce the power of
some tests, it is consistent with many real markets in which
share is concentrated at one end of the product spectrum.
.1
.2
We examined the impact when the entry brand attacks the
more or less desirable brands in a post hoc analysis.
.7' .3 . 7'
.S'
. S' .4
Analysis and Results
.5
.6
As expected, subjects had very similar responses to the
two brands positioned close together. Table 4 reports the
average percentage of subjects who would consider each of
a pair of brands. We present the results so that brands 1 and
2 are in the less considered (desirable) area and brands 5
and 6 are in the more considered area in the product space.
In the control condition, consideration set size averages 3.38
Attribute 1 of the possible 6 existing brands. In the entry treatment con-
ditions (in which two similar entry brands are introduced),
the average set size was largest (4.20) when brands 1 and 2
subjects (Appendix B), as shown by the pretest. Figure 3 were dominated by the entrants and smallest (3.56) when
presents brand positions in the two-dimensional case. brands 1 and 2 dominated the entrants.
Study 2 uses a between-subject design. In the control con- Here, the percentage of respondents who considered
dition, six existing brands are positioned in three pairs along brands I and 2 drops significantly when the brands are dom-
the efficient frontier in the two- or three-attribute product inated by entries compared with when they dominate en-
space. In a given entry treatment condition, two entry tries or when they are in the control condition (17% versus
brands in a pair are introduced with a brief statement indi- 26% or 30%, p < .05). Similarly, brands 5 and 6's consider-
cating that they are new brands introduced only recently. ation level also drops significantly when they are dominated
We used pairs of brands with similar attribute structures for by entries compared with when they dominate entries or are
two main reasons. First, it greatly simplifies the design. Sec- in the control condition (53% versus 77% or 78%, p < .05).
ond, in many markets, major manufacturers offer a full line Brands 1 and 2 keep the same percentage as in the control
of essentially similar products, producing several clusters condition when they are in the compromise positions after
of products with similar attribute configurations. the entries, but lose 6% when they are in the extreme posi-
To avoid measurement contamination, only consideration tions. Similarly, brands 5 and 6 lose 2% from their control
set is measured. The consideration set measure includes the condition level when they are in the compromise positions
following two items: "Which brand(s) would you like to and 5% when they are in the extreme positions. These re-
keep for further consideration?" and "Which brand(s) are sults are in line with those reported in Study 1. Interestingly,
acceptable to you?" the decrease in consideration is slightly greater for weak than
Nine questionnaires were used to randomize the possible for strong brands, with the notable exception of dominated
order effects. Each questionnaire consists of five product cat- situations, in which the more desirable brands (5 and 6) show
egories, a control condition, and four different treatment con- a substantial drop. Basically, the results suggest that it is

Table 4
CONSIDERATIONSET MEMBERSHIP (STUDY 2)

Entry Conditions

Control Condition 1.1 & 2* 2.1 & 2 3.5&6 4.5&6 5.1 &2 6.5&6
Brands (6 brands in 3 pairs) dominating dominated dominating dominated compromise compromise
1 and 2 30% 26% 17% 33% 24% 30% 24%
3 and 4 61% 63% 44% 56% 31% 57% 43%
5 and 6 78% 80% 71% 77% 53% 73% 76%
7 and 8 (new) 9% 78% 22% 93% 21% 49%
Total size 3.38 3.56 4.20 3.76 4.02 3.62 3.84
N= 220 148 145 145 148 148 146
Note: Data are aggregated across six product categories.
*Brand's numbers correspond to those in Appendix B.
Context Effects, New Brand Entry, and Consideration Sets 371

Table 5 To test the impact of the target brands' desirability on

THE IMPACT OF NEW ENTRY ON CONSIDERATION SETS: the entry effects, the sample was split into two on the basis
REGRESSION ANALYSIS (STUDY 2) of whether the entry attacked the more or less desirable area.
The two subsamples show significantly different effects
Entry attacking Entry attacking Existing (p < .001). Several interesting results appear. First, the domi-
Independent Aggregate less considered more considered brands nated effect is weaker when the entry attack is in the less
variable results existing brands existing brands only<'
desirable area (difference significant at p < .01). Second, the
Intercept .31 .18 .42 .23 extreme effect is negative when the attack is in the less de-
(4.53)* (2.04)t (4.69)* (2.63)*
Compromise -.01 -.01 .01 -.00 sirable area but is slightly (but not significantly) positive
(-.31) (-.57) (.45) (-.01) when the attack is in the more desirable area (difference sig-
Extreme -.03 -.08b .02b -.02 nificant at p < .05). Third, the compromise and asymmetri-
(-1.75):j: (-3.22)* (.78) (-1.21)
cally dominating effect is positive when the attack is in the
Dominating .17 .20 .15 .02
(11.54)* (11.13)* (8.02)* (.76) less desirable area and negative when the attack is in the
Dominated -.26 -.16a -.35 a -.17 more desirable area (difference significant at p < .01). Fi-
(-18.42)* (-8.86)* (-18.42)* (-8.40)* nally, the compromise and asymmetrically dominated effect
Compromise and .04 .23a -.17 a .06
asymmetric (2.01)t (9.71)* (---{).65)* (2.83)*
has a nonsignificant positive sign when the attack is in the
dominating less desirable area and a negative sign when the attack is in
Compromise and .16 .03a -.34a -.14 the more desirable area (difference significant at p < .01).
asymmetric (-8.61)* (1.05) (-13.04)* (-7.09)* Although many of these results appear intuitive (i.e., the neg-
dominated
ModelR2 .14 .25 .26 .13 ative dominated effect is stronger when the attack takes place
in the more desirable area and the negative effect of being
Note: Values in parentheses are t-statistics, * =p < .01; t =p < .05; :j: =
p <.1.
extreme disappears when a brand is extreme in the more
Data are aggregated across six product categories. desirable area), others require further research.
aThe two coefficients are significantly different at p < .01 level. The entry effects reported here do not vary on the basis
bThe two coefficients are significantly different at p < .05 level.
<Entry brands were excluded and only six existing brands after entry
of average product involvement levels. Hence, consideration
were included in this column. set membership may not be as sensitive to involvement as
choice (Ratneshwar, Shocker, and Stewart 1987).
fairly easy to decrease the likelihood of a brand being con- DISCUSSION
sidered (particularly by introducing a dominated entry) but
difficult to increase it because consumers have a tendency We investigate the impact of new brand entry on consid-
to limit the size of their consideration sets and considera- eration set membership. Two experiments (one within-
tion of the new entrant accounts for almost all the increase subject and one between-subject) involving a total of 332
in the size of the consideration set. subjects and 18 products provide credible empirical support
Similar to Study 1, regression analysis is used to summa- for the hypothesized effects. The results show that becom-
rize the results. The dependent variable is again a two-item ing dominated or extreme after entry reduces consideration
measure of consideration (r = .68). The independent varia- but that becoming dominating or compromise does not in-
bles are dominating, dominated, compromise, and extreme crease consideration, partly supporting HI and H2 . Also, as-
similation has a significant impact, supporting H3. More-
effects, coded as in Study 1. In addition, Study 2 introduces
over, assimilation helps weak brands and hurts strong
two new variables: compromise and asymmetrically domi-
brands. Interestingly, these results also occur for small (two-
nating and compromise and asymmetrically dominated.
existing brand) markets and are even somewhat stronger in
When a dominated (dominating) pair of entry brands attacks
small markets than in large markets, supporting H4 and H5 .
a pair of existing brands, the middle existing pair becomes
Also, where the entry brands attack seems to affect the entry
a compromise and asymmetrically dominating (dominated) effects, which calls for further research. Finally, a com-
pair. The data are pooled across subjects, product catego- promise position has a positive impact when the market is
ries, and brands, and the model includes subject and prod- small and a negative one when the number of brands in the
uct dummies to control for subject and product differences. market increases to the extent that it is hard to stand out in
The results show that the dominating, dominated, and ex- the crowd.
treme effects are as hypothesized and significant (Table 5), Of course, there are limitations to this study that make
supporting HI' H2 , and H3 . The dominating effect is posi- drawing strong conclusions unwise. This study employed
tive (p < .01), the dominated effect is negative (p < .01), the consumers in a paper-and-pencil task. Moreover, the brands
extreme is negative (p < .1), and the compromise effect is were not available for physical inspection, only described
not significant, consistent with the findings in Study 1. Also, on two or three attributes, and brand names were absent,
the compromise and asymmetrically dominating effect is pos- suggesting that the results would be weaker in other situa-
itive (p < .05), and the compromise and asymmetrically dom- tions. In addition, the design confounds the entry position
inated effect is negative (p < .01). Interestingly, an asym- with the brand's attribute values so no perfect control con-
metrically dominating effect more than offsets the negative dition is available. Further studies employing different
impact of a compromise position. measurement orders and methods (e.g., multiple measures
372 JOURNAL OF MARKETING RESEARCH, AUGUST 1994

of each construct) are clearly needed to substantiate these nesses, this study seems to provide encouraging support for
results as are studies that track the process (e.g., with con- focusing on changing consideration set membership as a pri-
current or retrospective protocols) by which entry affects mary impact of new brand entry. Obviously, further research
consideration, preference, and choice. Yet in spite of weak- is needed to corroborate and refine this conclusion.

Appendix A
TWO EXISTING BRAND MARKETS (STUDY 1)
(New entrants' positions are indicated by numbers corresponding to Figure 1)

Orange juice Cars Calculator battery 1Vset Restaurant Stockfund Papertowel Stereo speaker
mile picture price return relative
taste price per ride expected price quality durable food dinner rate risk sound price
Brands quality 64fo. galion quality life(hr) pair (lines) (month) quality for 2 annual 10= best absorb strength quality pair
existing:
A 70 $2.29 36 65 14 $1.75 375 48 78 $ 70 28% 6.5 92 62 91 $340
B* 90 $3.49 21 98 26 $3.25 254 82 95 $120 9% 9.5 70 88 75 $175
entries:
1 68 $2.39 34 62 13 $1.85 365 46 76 $ 73 27% 6.3 90 60 90 $355
l' 62 $2.69 28 53 10 $2.15 315 36 69 $ 82 23% 5.5 84 54 86 $395
2 68 $2.19 38 62 13 $1.65 385 46 76 $ 67 29% 6.3 94 60 92 $355
2' 62 $1.89 44 53 10 $1.35 435 36 69 $ 58 33% 5.5 100 54 95 $395
3 72 $2.19 38 68 15 $1.65 385 50 80 $ 67 29% 6.7 94 64 92 $325
3' 78 $1.89 44 77 18 $1.35 435 60 87 $ 58 33% 7.5 100 70 95 $285
4 72 $2.39 34 68 15 $1.85 365 50 80 $ 73 27% 6.7 90 64 90 $325
4' 78 $2.69 28 77 18 $2.15 315 60 87 $ 82 23% 7.5 84 70 86 $285
Note: *Brand B is the targetexistingbrand,andbrandnumber corresponds to that in Figure1.

(New entrants' positions are indicated by numbers corresponding to Figure 2)

Beer Apartment Walking shoes Camera

close to
price size campus durable lens
Brands 6-pack quality sqft (minute) price (month) quality price

existing:
1* $8.99 92 620 30 $33 12 61 $129
2 $8.49 90 590 27 $37 12 64 $149
3 $7.99 86 560 23 $43 15 68 $179
4 $6.99 81 520 19 $45 16 71 $229
5 $5.99 73 480 14 $52 20 77 $279
6 $4.99 64 450 11 $73 22 83 $319
7 $3.99 52 390 5 $84 26 91 $369
8 $2.99 44 360 2 $92 32 98 $409
entries:
1' $9.49 77 500 33 $48 10 58 $249
2' $9.49 95 650 33 $29 10 58 $109
3' $6.49 95 650 17 $29 18 74 $109
4' $6.49 77 500 17 $48 18 14 $249
Note: *Brand number corresponds to thatin Figure2.

Appendix B
SIX EXISTING BRAND MARKETS (STUDY 2)

Orange Juice Compsedan Light bulb PC Air-conditioner Stereo speakers

nutritional reliabi- mile usage price base speed cooling energy sound price power
Brands taste value lity p.g. life(hr) pair price memory (MHz) (btulhr) efficiency price quality pair (watts)

existing:
1* 76 83 80 39 1000 $1.25 $1600 4 25 5000 9.0 $275 89 $319 85
2 77 82 81 38 1100 $1.29 $1590 4 25 4900 8.9 $275 88 $315 80
3 83 75 85 35 1400 $1.45 $1400 2 20 5500 8.5 $300 83 $299 75
4 84 74 86 34 1500 $1.49 $1390 2 20 5400 8.4 $300 82 $295 70
5 90 67 90 31 1800 $1.65 $1200 1 16 6000 8.0 $325 77 $279 65
6 91 66 91 30 1900 $1.69 $1190 1 16 5900 7.9 $325 76 $275 60
Context Effects, New Brand Entry, and Consideration Sets 373

Appendix B-(Continued)

Orange Juice Compsedan Light bulb PC Air-conditioner Stereo speakers

nutritional reliabi- mile usage price base speed cooling energy sound price power
Brands taste value lity p.g. life(hr) pair price memory (MHz) (btulhr) efficiency price quality pair (watts)

entries:
7a 83 91 85 42 1400 $1.05 $1400 6 32 5500 9.5 $250 95 $299 95
8a 84 90 86 41 1500 $1.09 $1390 6 32 5400 9.4 $250 94 $295 90
7b 69 75 75 35 600 $1.45 $1800 2 20 4500 8.5 $300 83 $339 75
8b 70 74 76 34 500 $1.49 $1790 2 20 4400 8.4 $300 82 $335 70
7C 97 75 95 35 2200 $1.45 $1010 2 20 6500 8.5 $300 83 $259 75
8c 98 74 96 34 2300 $1.49 $1000 2 20 6400 8.4 $300 82 $255 70
7d 83 55 85 27 1400 $1.85 $1400 .8 12 5500 7.5 $350 71 $299 55
8d 84 54 86 26 1500 $1.89 $1390 .8 12 5400 7.4 $350 70 $295 50
» 97 55 95 27 2200 $1.85 $1010 .8 12 6500 7.5 $350 71 $259 55
s- 98 54 96 26 2300 $1.89 $1000 .8 12 6400 7.4 $350 70 $255 50
7f 69 91 75 42 600 $1.05 $1800 6 32 4500 9.5 $250 95 $339 95
8f 70 90 76 41 700 $1.09 $1790 6 32 4400 9.4 $250 94 $335 90
Note:"entriesdominate 1 and 2. bentries are dominated by 1 and 2. centries dominate 5 and 6. dentries are dominatedby 5 and 6. e5 and 6 are compromise.
fl and 2 are compromise.
*Brandnumbercorrespondsto that in Figure 3.

REFERENCES ularity and the Similarity Hypothesis," Journal of Consumer

Research, 9 (June), 90-98.
Alba, Joseph W. and J. Wesley Hutchinson (1987), "Dimensions
of Consumer Expertise," Journal of Consumer Research, 13 __ and Christopher Puto (1983), "Market Boundaries and Prod-
(March), 411-54. uct Choice: lliustrating Attraction and Substitution Effects," Jour-
Bettman, James R. (1979), An Information Processing Theory of
nal of Consumer Research, 10 (June), 31-44.
Consumer Choice. Reading, MA: Addison-Wesley. Kardes, Frank R., Paul M. Herr, and Deborah Marlino (1989),
__ and C. Whan Park (1980), "Effects of Prior Knowledge and "Some New Light on Substitution and Attraction Effects," Ad-
Experience and Phase of the Choice Process on Consumer De- vances in Consumer Research, 16,203-208.
cision Processes: A Protocol Analysis," Journal of Consumer __ and Gurumurthy Kalyanaram (1992), "Order-of-Entry Effects
Research, 7 (December), 234-48. on Consumer Memory and Judgment: An Information Integra-
Biehal, Gabriel and Dipankar Chakravarti (1986), "Consumers' Use tion Perspective," Journal ofMarketing Research, 29 (August),
of Memory and External Information in Choice: Macro and 343-57.
Micro Perspective," Journal of Consumer Research, 12 (March), ----' ----' Murali Chandrashekaran, and Ronald J. Domoff (1993),
382-405. "Brand Retrieval, Consideration Set Composition, Consumer
Brisoux, Jacques E. and Michel Laroche (1981), "Evoked Set For- Choice, and the Pioneering Advantage," Journal of Consumer
mation and Composition: An Empirical Investigation under a Research, 20 (June), 62-75.
Routinized Response Behavior Situation," Advances in Con- Mahajan, Vijay, Eitan Muller, and Frank M. Bass (1990), "New
sumer Research, 8, 357-61. Product Diffusion Models in Marketing," Journal of Market-
Carpenter, Gregory S. and Kent Nakamoto (1989), "Consumer Pref- ing, 54 (January), 1-26.
erence Formation and Pioneering Advantage," Journal of Mar- Mishra, Sanjay, U.N. Umesh, and Donald E. Stem (1993), "Ante-
keting Research, 26 (August), 285-98. cedents of the Attraction Effect: An Information-Processing Ap-
Cohen, Joel B. and Kunal Basu (1987), "Alternative Models of proach," Journal of Marketing Research, 30 (August), 331-49.
Categorization: Toward a Contingent Processing Framework," Narayana, Chern L. and Ron J. Markin (1975), "Consumer Be-
Journal of Consumer Research, 13 (March), 455-72. havior and Product Performance: An Alternative Conceptuali-
Farquhar, Peter H and Anthony R. Pratkanis (1987), "Phantom zation," Journal of Marketing, 39 (October), 1-6.
Choices: The Effects of Unavailable Alternatives on Decision Nedungadi, Prakash (1990), "Recall and Consumer Consideration
Making," working paper, Carnegie Mellon University. Sets: Influencing Choice Without Altering Brand Evaluations,"
Gensch, Dennis H. (1987), "A Two-Stage Disaggregate Attribute Journal of Consumer Research, 17 (December), 263-76.
Choice Model," Marketing Science, 6 (Summer), 223-39. Pan, Yigang and Donald R. Lehmann (1993), "The Influence of
Glazer, Rashi, Barbara E. Kahn, and William L. Moore (1991), New Brand Entry on Subjective Brand Judgments," Journal of
"The Influence of External Constraints on Brand Choice: The Consumer Research, 20 (June), 76-86.
Lone-Alternative Effect," Journal of Consumer Research, 18 Ratneshwar, Srinivasan, Allan D. Shocker, and David W. Stewart
(June), 119-25. (1987), "Toward Understanding the Attraction Effect: The Im-
Hauser, John R. and Steven M Shugan (1983), "Defensive Mar- plications of Product Stimulus Meaningfulness and Familiarity,"
keting Strategies," Marketing Science, 2 (Fall), 319-60. Journal of Consumer Research, 13 (March), 520-33.
__ and Birger Wernerfelt (1990), "An Evaluation Cost Model Roberts, John H. and James M. Lattin (1991), "Development and
of Consideration Sets," Journal of Marketing Research, 27 Testing of a Model of Consideration Set Composition," Jour-
(March), 393-408. nal of Marketing Research, 28 (November), 429-40.
Howard, John A. and Jagdish N. Sheth (1969), The Theory ofBuyer Rumelhart, David E. and Donald A. Norman (1972), "Accretion,
Behavior. New York: John Wiley & Sons, Inc. TIming and Restructuring: Three Models of Learning," in Or-
Huber, Joel, John W. Payne, and Christopher Puto (1982), "Add- ganization of Memory, E. Tulving and W. Donaldson, eds. New
ing Asymmetrically Dominated Alternatives: Violations of Reg- York: Academic Press, Inc., 197-246.
374 JOURNAL OF MARKETING RESEARCH, AUGUST 1994

Shocker, Allan D., Moshe Ben-Akiva, Bruno Boccara, and Prakash _ (1991), "The Effect of Buying Decisions on Consumers' As-
Nedungadi (1991), "Consideration Set Influences on Customer sessment of Their Tastes," Marketing Letters, 2 (January), 5-
Decision-Making and Choice: Issues, Models and Suggestions," 14.
Marketing Letters, 2 (August), 181-98. _ and Amos Tversky (1992), "Choice in Context: Tradeoff Con-
trast and Extremeness Aversion," Journal of Marketing Re-
Shugan, Steven M. (1980), "The Cost of Thinking," Journal of
search, 29 (August), 281-95.
Consumer Research, 7 (September), 99-111.
Sujan, Mita and James R. Bettman (1989), "The Effects of Brand
Simonson, Itamar (1989), "Choice Based on Reasons: The Case Positioning Strategies on Consumers' Brand and Category Per-
of Attraction And Compromise Effects," Journal of Consumer ceptions: Some Insights From Schema Research," Journal of
Research, 16 (September), 158-74. Marketing Research, 26 (November), 454-67.

• You can have

it all...
What you're reading now plus
the important articles in 800 other business and
management magazines, in a matter of
minutes.

The ABIIINFORM™ businessdatabase and your

computer give you access to article summaries
from magazines worldwide.

Call the publishers of ABIIINFORM at 800/626-

2823, today: