Content uploaded by L. J. Shrum
Author content
All content in this area was uploaded by L. J. Shrum on Jul 04, 2019
Content may be subject to copyright.
Undermining the Restorative Potential of
Compensatory Consumption: A Product’s
Explicit Identity Connection Impedes
Self-Repair
NIMISH RUSTAGI
L. J. SHRUM
When people experience threats to important aspects of their self-concept (e.g.,
power, intelligence, sociability), they often compensate by consuming products
that symbolize success, mastery, or competence on the threatened self-domain
(within-domain compensatory consumption). Our research examines whether
such compensatory consumption is effective in repairing the self-concept. Across
seven experiments, we show that whether compensatory consumption is effective
depends on the extent to which the connection between the compensatory prod-
ucts and the threatened domains is made explicit. When the connections are
made explicit (e.g., through product names and marketing slogans), self-repair is
impeded, but when the connections are only implicit (product is inherently sym-
bolic of self-threat domain), self-repair can be successful. We further show that
these differential effects of product connection explicitness are mediated by rumi-
nation: explicit connections induce rumination about the self-threat, which under-
mines self-repair, whereas implicit connections cause no rumination, facilitating
self-repair. Our research provides a reconciliation of conflicting findings on self-
repair in previous research, and also shows that despite the differences in
efficacy, consumers compensate regardless of whether product connections are
implicit or explicit, which has implications for consumer well-being.
Keywords: self-concept, identity, compensatory consumption, materialism,
self-threat, self-discrepancy
People routinely experience setbacks in daily life, and
often these setbacks can be distinctly personal: being
passed over for a promotion at work, having a journal sub-
mission rejected, or being turned down for membership in
an exclusive club. The setbacks are particularly personal
because they may represent threats to fundamental aspects
of self-identity—in the previous examples, threats to
power, intelligence, and belongingness. Not surprisingly,
such self-threats produce aversive states from which people
are motivated to escape (Rucker and Galinsky 2013).
Although there are multiple avenues available for coping
with a damaged self-identity, one route is through symbolic
consumption (Wicklund and Gollwitzer 1982). Possessions
can serve to signal, either to the self or to others,
success, competence, or mastery of a self-identity domain
Nimish Rustagi (nimish.rustagi@gmail.com) is Deputy Press
Secretary to the President of India, New Delhi, India. L. J. Shrum
(shrum@hec.fr) is a professor of marketing, HEC Paris, Jouy-en-Josas,
France. Please address correspondence to L. J. Shrum. This research was
supported by a research grant from the HEC Foundation of HEC Paris and
Investissements d’Avenir (ANR-11-IDEX-0003/Labex Ecodec/ANR-11-
LABX-0047) awarded to the second author. This article is based on a por-
tion of the doctoral dissertation of the first author under the supervision of
the second author. The authors thank the editor, associate editor, and
reviewers for their helpful input and guidance through the review process.
The authors also thank Tina Lowrey, and participants in seminars at Tel-
Aviv University, London School of Economics, Hong Kong University,
City University of Hong Kong, and the HEC-INSEAD-ESSEC marketing
camp, for their helpful comments on this research. Supplementary materi-
als are included in the web appendix accompanying the online version of
this article.
Darren W. Dahl and Linda L. Price served as editors and Derek D.
Rucker served as associate editor for this article.
Advance Access publication July 28, 2018
V
CThe Author(s) 2018. Published by Oxford University Press on behalf of Journal of Consumer Research, Inc.
All rights reserved. For permissions, please e-mail: journals.permissions@oup.com Vol. 0 2018
DOI: 10.1093/jcr/ucy064
1
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
(Mandel et al. 2017;Rucker and Galinsky 2013), and pro-
vide a means for individuals to cope with self-threats (Gao,
Wheeler, and Shiv 2009;Rucker, Galinsky, and Dubois
2012). Thus, individuals experiencing a self-threat may be
drawn to products that are symbolic of the self-threat do-
main. Such processes may explain why Janice, a manager
who had her project funding abruptly curtailed by her boss,
is suddenly very enamored of the Aston Martin billboard ad
she passes every day, or why Bill, a student who received a
bad grade on his physics exam, picks up the Economist at
the convenience store rather than his usual GQ.
The question we address in this research is whether such
symbolic compensatory consumption works. That is, does
compensatory consumption help in restoring an individu-
al’s self-identity on the threatened domain? Although a
large literature has established that consumers routinely en-
gage in compensatory consumption behaviors in response
to self-threats (for reviews, see Mandel et al. 2017;Rucker
and Galinsky 2013), research addressing its efficacy is
both sparse and inconsistent. In this research, we test the
proposition that whether compensatory consumption
restores the damaged aspect of self depends at least in part
on contextual factors associated with the product.
Specifically, we posit that the efficacy of compensatory
consumption depends on the extent to which the compen-
satory product is explicitly connected to the self-threat.
Most research on compensatory consumption assumes (and
typically verifies via pretests) that a particular product is
inherently (implicitly) associated with self-domains such
as power, intelligence, belonging, and so forth. For exam-
ple, fountain pens (Gao et al. 2009) and dictionaries (Kim
and Rucker 2012) are considered to be inherently symbolic
of intelligence. However, the connections between a prod-
uct and a self-domain can also be made explicit, particu-
larly through marketing tactics. Examples include
compensating for threats to competence by buying a prod-
uct promoted as “A Skillful Board Game” (Lisjak et al.
2015, 1192), and compensating for threats to intelligence
with a pen marketed as a “Brain Pen” (Kim and Gal 2014,
532).
Across seven experiments, we demonstrate that whether
successful self-repair follows from compensatory con-
sumption depends on the extent to which the compensatory
product’s inherent symbolic linkage to the threatened self-
domain aspect is made explicit. We show that compensa-
tion for self-threats with products that are implicitly con-
nected to the self-threat leads to successful self-repair.
However, when the compensatory products are explicitly
connected to the threatened identity domain, self-repair is
thwarted. We further show that these differential self-
repair outcomes are mediated by rumination about the self-
threat: explicit connections induce rumination about the
self-threat, which impedes self-repair, whereas implicit
connections do not cause rumination, thus facilitating
self-repair. We also demonstrate an important boundary
condition to these effects by showing that when the self-
threats themselves are implicit (do not directly or explicitly
threaten a self-concept domain), explicit connections can
also facilitate self-repair. Importantly, we show that con-
sumers are apparently unaware of these differential effects,
as they show the same compensatory consumption prefer-
ences regardless of whether the symbolic connections to
the self are implicit or explicit. Thus, using the examples
noted earlier, whether the thought of an Aston Martin suc-
cessfully restores Janice’s sense of power may depend on
whether the billboard for Aston Martin that she notices
includes its tagline, “Power, Beauty, and Soul,” and
whether purchasing the Economist restores Bill’s sense of
intelligence may depend on whether he notices or remem-
bers its promotion as “A Gymnasium for the Mind.”
SELF-THREATS AND COMPENSATORY
CONSUMPTION
People are highly motivated to maintain a positive and
stable self-concept (Leary et al. 1995). Thus, when people
experience threats to important aspects of the self, particu-
larly core aspects such as self-esteem, belonging, efficacy
(power, control), and meaningful existence (Williams
2007), it creates discrepancies between actual and ideal
self-concepts, resulting in uncomfortable, aversive states
that people are motivated to remedy by bolstering or boost-
ing the particular aspect of the self that is threatened
(Higgins 1987;Kim and Rucker 2012). Although there are
multiple avenues available for resolving these self-
discrepancies (e.g., self-improvement, affirming the self in
a different domain, distraction; for a review, see Mandel
et al. 2017), one strategy is to engage in symbolic behav-
iors that signal mastery or competence on the threatened
self-domain (symbolic self-completion;Wicklund and
Gollwitzer 1982). Moreover, such symbolic signaling can
be accomplished through products and possessions. Thus,
when people experience threats to important aspects of
their self-identity, they may compensate by acquiring, dis-
playing, or even simply thinking about products that signal
success on the threatened domain, which Rucker and col-
leagues refer to as within-domain compensatory consump-
tion (Lisjak et al. 2015;Mandel et al. 2017).
1
Compensatory consumption has been demonstrated
across a broad spectrum of self-threats (Mandel et al.
2017). For example, relative to nonthreatened participants,
participants whose power was threatened were willing to
pay more for high-status products (Rucker and Galinsky
1 We adopt Rucker and Galinksy’s (2013, 207) definition of compen-
satory consumption as “the desire for, acquisition, or use of products
to respond to a psychological need or deficit.” Thus, compensatory
consumption captures a broad variety of phenomena beyond actual
consumption.
2JOURNAL OF CONSUMER RESEARCH
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
2008,2009) and preferred products with bigger status lo-
gos (Lee and Shrum 2012;Rucker and Galinsky 2009),
those whose intelligence was threatened were more likely
to choose and were willing to pay more for products signal-
ing intelligence (Gao et al. 2009), and those who were so-
cially excluded expressed greater preferences for products
signaling affiliation (Mead et al. 2011). Similar findings
have been observed for self-threats related to one’s self-
esteem (Lee and Shrum 2012;Lee, Shrum, and Yi 2017),
personal freedom (Levav and Zhu 2009), and social system
(Cutright et al. 2011).
The Efficacy of Compensatory Consumption
Despite considerable evidence that consumers use pos-
sessions to compensate for self-threats, one question that
has received relatively little attention is whether such com-
pensatory consumption is actually successful in restoring
the threatened self-concept. Although little, if any, research
has specifically investigated the efficacy of compensatory
consumption, at least two published studies have reported
evidence of self-identity repair following compensatory
consumption. Gao et al. (2009, experiments 1 and 3) first
threatened participants on a self-domain (e.g., intelligence,
exciting person) by manipulating their confidence in these
traits, and then manipulated the opportunity to symboli-
cally compensate with products by having participants ei-
ther choose from a list of products that they could win in a
lottery (experiment 1) or that were ostensibly to be used as
prizes in future studies (experiment 3) that were either as-
sociated or not associated with the self-threat domain.
Their findings showed that compensating for the self-threat
with products symbolically related to the self-threat do-
main restored the self-ratings of threatened individuals to
the same level as those who were not threatened. These
findings are consistent with research showing that the need
for self-expression through brands is finite and can be sati-
ated through identity-related consumption (Chernev,
Hamilton, and Gal 2011).
Although the Gao et al. (2009) findings show that com-
pensatory consumption may effectively satiate the need to
bolster self-identity following a self-threat, other research
suggests that it may not. Lisjak et al. (2015) reported the
results of a series of experiments that investigated the
effects of compensatory consumption on self-control. They
found that compensating with products that signal success
on the self-threat domain (within-domain compensatory
compensation) resulted in increased rumination about the
self-threat, which in turn reduced self-control on subse-
quent tasks, compared to threatened participants who com-
pensated with products that signaled success on a self-
domain unrelated to the threat (across-domain compensa-
tion). Presumably, such within-domain compensation
served to keep the previous threat or failure active in mem-
ory, cuing ruminative thoughts.
Although Lisjak at al. (2015)’s research does not directly
address self-repair, the finding that (within-domain) com-
pensatory consumption increases rumination about the
self-threat strongly suggests that such compensatory con-
sumption should impede self-repair. Self-threatened indi-
viduals engage in compensatory consumption to escape an
aversive state by seeking out products that signal success
on the threatened domain. However, Lisjak et al.’s research
suggests that such compensation can have the insidious ef-
fect of keeping the aversive state top-of-mind. In other
words, within-domain compensatory consumption may ac-
tually prolong the aversive state from which threatened
individuals are trying to escape. Even worse, if those indi-
viduals are unaware of these effects, they may actually
continue to seek out compensatory products that not only
do not fulfill their goal, but make it worse. These findings
seemingly run counter to the findings of Gao et al. (2009),
who demonstrated self-repair through within-domain com-
pensatory consumption.
Explaining the Discrepancies. The discrepancies just
noted raise the question of how to explain them. One possi-
bility, which provides the motivation for this research,
relates to differences in the extent to which the connections
between the compensatory product and the self-threat do-
main are made explicit. As noted earlier, researchers test-
ing compensatory consumption typically use products that
are pretested for their inherent symbolic connections to the
self-threat domain, and this was true for the research
reported by Gao et al. (2009), which demonstrated self-
repair through compensatory consumption. Importantly,
the implicit symbolic links are evidenced by the fact that
these products are not presented with any explicit informa-
tion (marketing slogans, brand names) that connects them
to the threatened self-domain. We refer to these examples
as implicit connections because the connection is not
overtly referenced but rather inferred from the inherently
symbolic connections between the product and the self-
concept domain.
2
In contrast, in the research reported by Lisjak et al.
(2015), which focused on the effects of within- versus
across-domain compensatory consumption, the symbolic
connections between the compensatory products and the
self-threat domain were made explicit. For example, partic-
ipants who were threatened on the domain of competence
were given the opportunity to compensate with products
such as a book titled Genius: Simple Ways to Become
Instantly Smarter, and participants who were threatened on
2 Here (implicit connections) and later (implicit threats, experiment
5) we use the term implicit to indicate the lack of any overt or explicit
reference to a self-threat domain. Our usage is conceptually similar to
the distinction between explicit attitude measures (direct, overt mea-
sure of the construct) and implicit attitude measures (inferred from
other properties such as accessibility). Bhattacharjee, Berger, and
Menon (2014) also use the term explicit to refer to overt references to
identity domains.
RUSTAGI AND SHRUM 3
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
the domain of sociability were asked to write about a time
in which they had played a “Social Board Game” (Lisjak
et al. 2015, 1192) that made them feel connected to their
friends and happy with their relationships. We refer to
these examples as explicit connections because aspects of
the product—in this case the product name—explicitly ref-
erence the self-threat domain. The results showed that
compensation with these explicitly connected products
resulted in rumination about the self-threat, which arguably
should impede self-repair.
Given these findings, it is possible that the explicit con-
nections between the self-threat domain and the compensa-
tory products is what is driving rumination about the
self-threat, rather than within-domain compensation more
generally. If so, then explicit connections should impede
self-repair, but implicit connections should not. Our re-
search tests these possibilities.
Explicit versus Implicit Connections and Compensatory
Consumption. Our reasoning regarding the distinction be-
tween explicitly versus implicitly connected compensatory
products raises additional questions about compensatory
consumption more generally. For example, do threatened
individuals compensate for self-threats regardless of
whether the compensatory products are explicitly or im-
plicitly connected to the self-threat? Moreover, if they do,
do they prefer one over the other? There are at least three
possibilities. First, given that people are likely unaware of
the connection between an experienced self-threat and their
preferences for products that signal success on the threat-
ened dimension, they should compensate regardless of
whether the compensatory products are implicitly or ex-
plicitly connected, consistent with previous research
(Bhattacharjee et al. 2014;Kim and Gal 2014;Rucker and
Galinsky 2008,2009). A second possibility is that threat-
ened individuals may compensate with both implicitly and
explicitly connected products, but the strength of the
effects may be greater for explicitly connected products be-
cause the connections are more overt. A third possibility is
that individuals will compensate through implicitly con-
nected products but not explicitly connected products.
Despite the studies just noted showing compensatory con-
sumption with explicitly connected products, it is also rea-
sonable to think that explicit connections may weaken or
eliminate the compensatory consumption effects because
they draw conscious attention to the relation between the
product and the recent self-threat, which may result in
mental processes similar to those underlying contrast
effects in priming research (Martin 1986;Wyer and Srull
1989). Our research also tests these possibilities.
Explicit Connections, Explicit Threats, and
Underlying Processes
Explicit versus Implicit Product Connections. We have
proposed that whether compensatory consumption success-
fully repairs a damaged self-concept depends on the extent
to which the compensatory product is explicitly connected
to the self-threat domain. To reiterate, we consider a connec-
tion to be explicit if aspects of the product explicitly refer-
ence the self-threat domain. For example, a potential
compensatory product with the slogan “The intelligent
choice” is unambiguously explicitly linked to an intelligence
threat, as the slogan uses the exact same word as the self-
threat. Similarly, a product labeled as “the smart choice”
(intelligence ¼smart) or “the brain pen” (intelligence ¼brain)
should also provide an explicit link, but in these cases the slo-
gans and self-threats do not use identical descriptors, but ones
that are very closely related semantically (synonyms). In con-
trast, providing a simple choice of a pen would be an example
in which the potential compensatory product is implicitly con-
nected to the self-threat, because pens are inherently symbolic
of intelligence (Gao et al. 2009).
These examples of explicitly and implicitly connected
products can be viewed as representing opposite ends of an
explicitness continuum. Even in the examples in which the
links are between the explicit self-threat (e.g., intelligence)
and synonyms of intelligence (smart, brainy), the semantic
links are likely so strong that there is little difference in the
strength of activation between exact words and its syno-
nyms. In other cases, however, semantic links may exist
that are arguably weaker. Consider a potential compensa-
tory product for a power threat that uses a slogan that refer-
ences status. In this case, the level of explicitness rests on
the extent to which status and power are sufficiently
closely linked in memory that activation of one concept
(status) will activate the other concept (power). In such
cases, the level of activation may be less than when more
explicit links (in command, power) are used. If so, then
less rumination should occur when the links are less ex-
plicit than when they are more explicit. However, to the
extent that the less explicit links are sufficient to reactivate
the self-threat domain, less explicit links should result in
more rumination than when the links are only implicit.
Explicit versus Implicit Threats. The general process we
have outlined is one in which an aspect of a consumer’s self-
concept is threatened, creating a discrepancy between the ac-
tual and ideal selves, which activates a goal to reduce the
self-discrepancy (Higgins 1987;Mandel et al. 2017). This
process assumes that the self-threat (and corresponding goal)
is explicitly stored in memory, and then reactivated when
consumers consider products that are explicitly connected to
the self-threat. For example, when an individual receives a
poor score on a standard intelligence test, the negative perfor-
mance has overt connections to failure on the intelligence
4JOURNAL OF CONSUMER RESEARCH
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
domain. Likewise, typical experimental manipulations of
self-threats ask participants to recall and write about a time in
which someone had power over them (Rucker and Galinsky
2008),askthemtorecallatimeinwhichtheyhadnotfeltas
competent as they would have desired (Lisjak et al. 2015), or
give participants bogus feedback that they performed poorly
on a cognitive intelligence test (Kim and Gal 2014). In these
cases, the self-threat is clear and explicit. Participants in a
study may not be aware of the purpose of such experimental
manipulations or immediately assess the impact of the threat
on their self-concept. However, when an individual facing an
explicit threat subsequently compensates with an explicitly
connected product, the product’s explicit references to the
threatened domain should activate the failure-related cues
contained in the explicit threat experience, resulting in threat-
related rumination.
However, suppose that the threat itself is not an explicit
one, but rather one that is subtler and more implicit.
Consider the research reported by Gao et al. (2009), which
showed that self-threat deficits may occur in the absence of
explicit references to a threat domain. In their experiments,
rather than explicitly implementing a self-threat, they
merely undermined confidence in a self-concept domain.
Specifically, rather than having participants recall a time in
which they did not feel as intelligent, competent, or excit-
ing as they would desire, all participants listed personal
characteristics that suggested they were intelligent (experi-
ments 1 and 2), competent, or exciting (experiment 3).
However, prior to these tasks, which actually mimic self-
concept boosts, the researchers manipulated participants’
general self-confidence by having them write with their
nondominant hand or list experiences in which they felt
doubt or uncertainty. Importantly, the confidence threat
was general, not specific to any particular self-concept do-
main, and thus participants compensated in response to a
general self-concept threat (Gao et al. 2009). In these
cases, given that there were no explicit and specific threats
to participants’ self-concepts, there should not be a specific
self-threat stored in memory. If so, explicitly connected
compensatory products may also result in self-repair, be-
cause no explicit threat exists in memory to be activated,
and consequently there is no specific threat to be ruminated
upon. Our research tests these possibilities.
Summarizing, Lisjak et al. (2015) presented evidence that
within-domain compensatory consumption may undermine
self-control, and by extension, self-repair. They proposed
that the symbolic relation between the compensatory prod-
uct (e.g., a pen) and the self-threat domain (e.g., intelli-
gence) may reactivate the self-concept domain that had been
previously threatened, which cued ruminative thoughts
about the self-concept deficit. In their studies, they further
showed that the process of rumination inhibited self-control.
We are suggesting a slightly different process: it is not the
symbolic connection per se that reactivates the self-threat
and induces rumination, but rather that self-threat
reactivation occurs when the connections between the com-
pensatory product and the self-threat domain are explicit,
which was the case in the compensatory stimuli used by
Lisjak et al. (2015). If our reasoning is correct, it provides a
compelling reconciliation of the findings of Lisjak et al.
(2015) and Gao et al. (2009), the latter of whom documented
self-repair through within-domain compensatory consump-
tion with implicitly connected products.
Specifically, we expect that 1) threatened individuals
will show a preference for products symbolically related to
self-threat domain, regardless of whether the products’
connection to the threatened identity domain is implicit or
explicit, 2) compensating with implicitly connected prod-
ucts will result in self-repair to the damaged self-identity
domain, but compensating with explicitly connected prod-
ucts will impede self-repair, 3) these differential effects of
compensating with implicitly versus explicitly connected
products on self-repair will be mediated by rumination,
and 4) when the self-threats themselves are implicit, com-
pensation via explicitly connected products will also result
in self-repair. The general model is shown in figure 1.
OVERVIEW OF EXPERIMENTS
We tested our hypotheses in seven experiments using a
multistep process that varied the domain of the self-threat
(power, intelligence, sociability), the compensation process
(voluntary, forced), the explicitness of the connections, the
method to assess self-repair (pre- vs. post-compensation
measures, sequential compensation measures), and sample
composition (US MTurk participants, Indian college stu-
dents, French master’s students). Experiment 1 tested the
self-repair hypothesis. Experiment 2 provided a conceptual
replication of the self-repair hypothesis and also tested the
hypothesis that compensatory consumption will occur re-
gardless of whether the compensatory products are implic-
itly or explicitly connected to the self-threat. Experiments
3a, 3b, and 4 tested the hypothesis that the differential
effects of compensatory consumption via implicitly versus
explicitly connected products is mediated by rumination.
Finally, experiment 5 and web appendix experiment 1
tested a boundary condition of the hypothesis that explic-
itly connected compensatory products impede self-repair—
namely, that when the self-threats are implicit, explicitly
connected products will also facilitate self-concept repair.
In all experiments, we analyzed the data only after all
measures had been collected. We excluded participants
based on a priori rules (described within each study) that
were applied before any data analyses. We included mood
as a covariate in all studies because self-threats have been
shown to affect mood in previous research (Duclos, Wan,
and Jiang 2013;Vohs and Heatherton 2004). All partici-
pants provided informed consent. All measures and manip-
ulations are provided in the web appendix.
RUSTAGI AND SHRUM 5
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
EXPERIMENT 1: EXPLICIT VERSUS
IMPLICIT CONNECTION AND
SELF-REPAIR (INTELLIGENCE THREAT)
Experiment 1 tested the hypothesis that compensating
with implicitly connected products leads to self-repair, but
compensating with explicitly connected products does not.
We either threatened or did not threaten participants on the
domain of intelligence, and then gave them the opportunity
to compensate for the self-threat but manipulated whether
they compensated via products with either implicit or ex-
plicit connections to the self-threat. We then assessed
whether this initial compensation opportunity would lead
to self-repair by providing participants with a second op-
portunity to compensate for the intelligence threat by
assessing their relative preference for an intelligence-
related product compared to a product unrelated to
intelligence. If the initial compensatory choice leads to
self-repair, and thus the need to bolster intelligence is sati-
ated, then threatened participants should not continue to
seek out symbols of intelligence, and thus should not dis-
play greater preferences for subsequent intelligence-related
products than should nonthreatened participants. However,
if compensatory consumption does not restore the self on
the domain of intelligence, then threatened participants
should continue to display greater preferences for subse-
quent intelligence-related products than should nonthreat-
ened participants.
Method
Participants, Design, and Procedure. Five hundred
eight members of the Amazon Mechanical Turk (MTurk)
online panel (M
age
¼33.2 years; 321 women) participated
FIGURE 1
THEORETICAL MODEL OF THE COMPENSATORY AND SELF-REPAIR PROCESS
Explicit self-threat
Implicit self-threat
Self-threat
(Explicit)
Compensatory
consumpon
(Explicitly connected products)
Ruminaon
Compensatory
consumpon
(Implicitly connected products)
Self-repair
+
+
+
-
Self-threat
(Implicit)
Compensatory
consumpon
(Explicitly connected products)
Compensatory
consumpon
(Implicitly connected products)
Self-repair
++
+
+
+
6JOURNAL OF CONSUMER RESEARCH
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
in return for a nominal fee and were randomly assigned to
conditions in a 2 (intelligence threat: yes vs. no) 2 (prod-
uct connection: explicit vs. implicit) between-subjects de-
sign. Participants were told that they would be
participating in several short studies. First, in a study osten-
sibly about understanding the different ways in which peo-
ple recount past experiences, participants were either
threatened on intelligence by writing about a time when
they had questioned their own intelligence or related abili-
ties (Lisjak et al. 2015) or not threatened (writing about
their last visit to a grocery store).
Next, we provided participants with an opportunity to
compensate using a choice task adapted from Gao et al.
(2009) and manipulated whether the connections between
the products and the self-threats used in the choice task
were implicit or explicit. Framed as a prize selection study,
the participants were asked for their help in identifying the
prizes that we could give to participants in our future stud-
ies. Participants were presented with three products, also
adapted from Gao et al. (2009), that naturally symbolize in-
telligence: Encyclopedia Britannica, the board game
Scrabble, and a subscription to National Geographic mag-
azine. In the explicitly connected product condition, each
of the products was presented along with a slogan that
made the product’s connection with intelligence explicit
(e.g., the Scrabble board game had the slogan “The smart-
est mind game,” with “smart” being explicitly connected to
intelligence), whereas in the implicitly connected condi-
tion, these same products were presented without any slo-
gans. Participants were asked to select one of the products
as their choice for the most suitable prize. This choice task
represented the initial compensatory consumption.
After participants completed a filler task that asked them
about their lifestyle preferences, which was included to dis-
guise the purpose of the study, participants completed an
ostensibly unrelated study on consumer preferences. They
were given two products and asked to indicate which they
preferred, one of which was related to intelligence (a book
entitled 1001 Inventions That Changed the World) and the
other unrelated to intelligence (a shirt), along a nine-point
scale (1 ¼surely prefer the shirt, 9 ¼surely prefer the
book). Order was counterbalanced. Next, participants com-
pleted a three-item mood scale (Cutright et al. 2011), and
scores on the three items were averaged to form a compos-
ite measure (a¼.86). Finally, participants provided demo-
graphic information and were asked about their
perceptions of the purpose of the study.
Results and Discussion
Exclusion Criteria. Data from 46 participants (9%)
were excluded from the analyses: 28 who did not write
anything or wrote nonrelevant content in the self-threat
task, 15 who guessed the purpose of the study, and three
who did not pass an attention check. The attention check
consisted of gender and age items that were measured at
the beginning and end of the study. Participants failed the
attention check if their answers did not match. The data
were analyzed for the remaining 462 participants.
Hypothesis Testing. We expected that threatened par-
ticipants compensating with explicitly connected products
would not experience self-repair, and thus would display
greater preferences for a product symbolic of intelligence
relative to nonthreatened participants. In contrast, we
expected that threatened participants who compensated
with implicitly connected products would experience self-
repair, and thus would not show greater preferences for the
intelligence-related product compared to nonthreatened
participants. To test these hypotheses, we conducted a 2
(threat: yes vs. no) 2 (product connection: explicit vs.
implicit) ANCOVA with preference for the intelligence-
related product as the dependent variable and mood as the
covariate. (The results did not materially differ when the
covariate was excluded from the analysis.) The results of
this analysis can be seen in figure 2.
There was a marginally significant main effect of threat
(F(1, 457) ¼3.17, p¼.076). More direct to our hypotheses,
the expected threat product connection interaction was
significant (F(1, 457) ¼4.46, p¼.035). Threatened partici-
pants who initially compensated via explicitly connected
products again demonstrated compensatory effects by dis-
playing a greater preference for intelligence-related prod-
ucts (M¼5.88, SE ¼.31) than did nonthreatened
participants (M¼4.73, SE ¼.28; F(1, 457) ¼7.71,
p¼.006). Thus, those who compensated for the intelli-
gence threat with explicitly connected products continued
to seek products symbolic of intelligence after compensa-
tory consumption, indicating no self-repair, consistent with
the findings of Lisjak et al. (2015). In contrast, threatened
participants who compensated with implicitly connected
products (M¼4.86, SE ¼.31) did not differ from non-
threatened participants (M¼4.96, SE ¼.28; F<1) on
preferences for intelligence-related products, indicating
that compensating with implicitly connected products
resulted in self-repair, consistent with the findings of Gao
et al. (2009).
The findings of experiment 1 provide support for our hy-
pothesis that compensatory compensation is successful in
repairing the self when the connection between the product
and threat domain is implicit. However, when the connec-
tion is made explicit, threatened participants continue to
seek threat-related products on a subsequent compensatory
opportunity, indicating a lack of self-repair. In experiment
2, to demonstrate the robustness of these effects, we pro-
vide a conceptual replication of experiment 1 in which we
threaten a different self-domain (power) and use a different
method of manipulating compensatory consumption that
also allows us to test the hypothesis that threatened individ-
uals will compensate for the threat regardless of whether
RUSTAGI AND SHRUM 7
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
the compensatory product is explicitly or implicitly con-
nected to the threat domain.
EXPERIMENT 2: EXPLICIT VERSUS
IMPLICIT CONNECTION AND SELF-
REPAIR (POWER THREAT)
The design of experiment 2 was similar to that of experi-
ment 1. We first threatened or did not threaten participants
on a self-relevant domain (power) and then gave them the
opportunity to compensate with products that were either
explicitly connected, implicitly connected, or unconnected
to status, followed by a second opportunity to compensate.
However, the method of the initial compensation was dif-
ferent from that of experiment 1. In experiment 1, the ini-
tial compensation was “forced,” in the sense that
participants could choose only products that were manipu-
lated to be threat-relevant. This procedure did not allow us
to test the hypothesis that threatened participants would
spontaneously compensate with symbolic products for both
explicitly and implicitly connected products. Thus, in ex-
periment 2, we measured participants’ initial compensation
by assessing their willingness to pay for products symboli-
cally related to the threat (status products).
We manipulated product connection using the same pro-
cedures as experiment 1. We also included a condition in
which the compensatory products were unconnected to the
threat domain to rule out the possibility that the self-threat
increased willingness to pay for any product. We expected
that threatened participants would be willing to pay more
for status products than would nonthreatened participants
for both explicitly and implicitly connected products, but
that willingness to pay for nonstatus products would not
differ by threat condition.
We next tested whether the initial compensatory con-
sumption would lead to self-repair by providing partici-
pants with a second opportunity to compensate for the
power threat by measuring their preference for additional
power-related choices (high-power vs. low-power job role,
importance of power-related job characteristics; Inesi et al.
2011). If compensatory consumption leads to self-repair
(satiation), then threatened participants should not display
greater preferences for subsequent power-related choices
than nonthreatened participants. However, if compensatory
consumption does not restore the self on the domain of
power, then the threatened participants should continue to
display greater preferences for subsequent power-related
choices than should nonthreatened participants. In addition,
threatened individuals in the unconnected condition, who
did not have the opportunity to compensate initially be-
cause the product was not symbolically connected to the
self-threat domain, should still seek to bolster their feelings
of power, and thus should be willing to pay more for
power-related choices than those who were not threatened.
This complex series of tests of compensatory consumption
provides a very conservative test of our hypotheses.
Method
Participants, Design, and Procedure. One hundred
forty-five members of the MTurk online panel
(M
age
¼39.7 years; 95 women) participated in return for a
nominal fee and were randomly assigned to conditions in a
2 (threat: yes vs. no) 3 (product connection: explicit vs.
implicit vs. none) between-subjects design. Participants
were told that they would be completing several short stud-
ies. In the first study, which used the same cover story as
experiment 1, participants were either threatened or not on
the self-domain of power. Those in the power-threat condi-
tion were asked to recall and write about an experience in
which someone else had power over them (Rucker and
Galinsky 2008), whereas those in the no-threat condition
were asked to write about the last time they went to the
grocery store.
Thereafter, as part of a separate study about willingness
to pay for different products, participants were either pre-
sented with two explicitly connected products, two implic-
itly connected products, or two unconnected (nonstatus)
products, and asked to indicate their willingness to pay for
each of the two. Based on previous research, we chose a
watch and a pen as status products (symbolic of power;
Rucker and Galinsky 2008) and a microwave oven and
chair as nonstatus products. Explicitly connected products
were accompanied by slogans that made their connection
to the domain of power explicit (e.g., a luxury watch car-
ried the slogan “For Those in Command, Every Moment,”
FIGURE 2
PREFERENCES FOR INTELLIGENCE-RELATED PRODUCT AS A
FUNCTION OF THREAT AND PRODUCT CONNECTION
(EXPERIMENT 1)
5.88
4.86
4.73 4.96
1
2
3
4
5
6
7
Explicitly connected Implicitly connected
Preference for
intelligence-related product
Threat No threat
8JOURNAL OF CONSUMER RESEARCH
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
with “in command” explicitly connected to power).
Implicitly connected products were the same status prod-
ucts but without the accompanying slogans. The uncon-
nected, nonstatus products also carried no slogan.
Participants then indicated their willingness to pay for these
products along a 12-point scale ranging from 10% of the retail
price to 120% of the retail price (Rucker and Galinsky 2008).
We averaged the willingness-to-pay scores for the two power-
related products (watch, status pen) to form a composite two-
item measure (r¼.58). Similarly, we averaged the
willingness-to-pay scores for the two non-threat-related prod-
ucts to form a composite two-item measure (r¼.64).
Next, as part of another ostensibly unrelated study about
job preferences, participants were asked to read two job
descriptions, adapted from Smith, Wigboldus, and
Dijksterhuis (2008), that differed on whether the job in-
cluded supervising others (high-power role) or being super-
vised by others (low-power role). (None of the descriptions
or items explicitly mention power; see web appendix.)
Participants then indicated which role they preferred on a
nine-point scale, with higher numbers indicating a greater
preference for the high-power role. Order of descriptions
was counterbalanced. Thereafter, as part of another study
to understand the characteristics of an ideal job, partici-
pants indicated the importance of job characteristics (Inesi
et al. 2011) that were either related to power (two items:
high position within the organization, control over key
resources) or unrelated to power (two items: friendly work-
ing atmosphere, interesting work), along a nine-point scale
(1 ¼not very important, 9 ¼extremely important). We
computed a composite measure of power-seeking by aver-
aging the three items measuring preference for and impor-
tance of power-related job characteristics (a¼.85) and a
composite measure of non-power-seeking by averaging the
two measures unrelated to power (r¼.58).
Finally, participants indicated their current mood with
the same three-item scale used in experiment 1 (a¼.87),
provided demographic information, and were asked about
their perceptions of the purpose of the study. Along with
mood, we also included gender as a covariate because the
items chosen for status products (watch, pen) may appeal
more to men than to women, and there is some evidence
that men have a greater need for power than do women
(Chusmir and Parker 1984).
Results and Discussion
Exclusion Criteria. Data from 13 participants (9%)
were excluded from the analyses (10 who guessed the pur-
pose of the study, two who did not write anything or wrote
nonrelevant content in the self-threat task, and one who in-
dicated he or she did not understand the willingness-to-pay
task), leaving 132 participants for analyses.
Hypothesis Testing: Compensatory Consumption. We
expected that participants who were threatened on power
would be willing to pay more for status products compared
to those who were not threatened, and that this would be
true regardless of whether the status products were explic-
itly or implicitly connected to the power domain. In con-
trast, we expected that willingness to pay for nonstatus
products would not differ by threat condition.
To test these hypotheses, we conducted a 2 (power
threat: yes vs. no) 3 (product connection: explicit vs. im-
plicit vs. none) analysis of covariance (ANCOVA), with
willingness to pay as the dependent variable and mood and
gender as covariates. (The results did not materially differ
when the covariates were excluded from the analyses.) The
results of this analysis can be seen in the top panel of
figure 3. There was a main effect of product connection
F(2, 124) ¼7.49, p¼.001) and a main effect of threat
(F(1, 124) ¼8.54, p¼.004); gender was marginally signif-
icant (F(1, 124) ¼3.66, p¼.058). As expected, threatened
participants were willing to pay more for the power-related
products than were nonthreatened participants for both ex-
plicitly connected (M
threat
¼5.05, SE ¼.52 vs. M
no threat-
¼3.36, SE ¼.49; F(1, 124) ¼5.64, p¼.02) and implicitly
connected products (M
threat
¼6.61, SE ¼.52 vs. M
no threat-
¼5.05, SE ¼.49; F(1, 124) ¼4.70, p¼.03), and these
effects did not differ (interaction: F<1). These results rep-
licate the findings of Kim and Gal (2014). In contrast, will-
ingness to pay for nonstatus products did not differ by
threat conditions (M
threat
¼6.16, SE ¼.51 vs. M
no threat-
¼5.75, SE ¼.49; F<1).
Although our hypotheses regarding compensation for
both explicitly and implicitly connected products was
strongly supported, as was our hypothesis that there would
be no differences in threat conditions for willingness to pay
for unconnected (non-status) products, the threat product
connection interaction was not significant (p¼.37). This
appears to be because even though the difference in will-
ingness to pay for non-status products between threat and
no-threat conditions was not close to significance (F<1),
it was nevertheless in the same direction as predicted for
explicitly and implicitly connected products, and appar-
ently of sufficient magnitude to render the interaction non-
significant. There was also an unexpected main effect of
product connection: willingness to pay was greater when
the connection was implicit than when it was explicit for
both threat and no-threat conditions. We interpret this find-
ing as indicating that participants found the product to be
less appealing when the slogan was included than when it
was not.
Hypothesis Testing: Self-Repair. We expected that
threatened participants compensating with explicitly con-
nected products would not experience self-repair, and thus
would continue to seek power, reflected by greater prefer-
ences for the high-power role and power-related job
RUSTAGI AND SHRUM 9
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
characteristics relative to nonthreatened participants. In
contrast, we expected that participants who compensated
with implicitly connected products would experience self-
repair, would consequently not continue to seek power,
and thus would show no preferences for the high-power
role and power-related job characteristics compared to
nonthreatened participants. In addition, we expected that
threatened participants exposed to nonstatus products, and
thus who had not previously had an opportunity to com-
pensate with threat-related products, would compensate for
the threat and express greater preferences for the power-
related items compared to nonthreatened participants.
To test these hypotheses, we conducted a 2 (power
threat: yes vs. no) 3 (product connection: explicit vs. im-
plicit vs. none) ANCOVA with power-seeking as the de-
pendent variable and mood and gender as covariates. The
results of this analysis can be seen in the bottom panel of
figure 3. (Some results were no longer significant when the
covariates were not included in the analyses. The results
without covariates are provided in the web appendix.) The
main effects of threat (F(1, 124) ¼5.91, p¼.02) and mood
(F(1, 124) ¼12.17, p¼.001) were significant, and gender
was marginally significant (F(1, 124) ¼3.24, p¼.074).
More central to our hypotheses, the threat product con-
nection interaction was marginally significant (F(2,
124) ¼2.42, p¼.093). As expected, threatened partici-
pants who initially compensated via explicitly connected
products again demonstrated compensatory effects by dis-
playing greater power-seeking (M¼6.73, SE ¼.49) com-
pared to nonthreatened participants (M¼4.84, SE ¼.46;
F(1, 124) ¼7.86, p¼.006). Thus, those who compensated
for the power threat with explicitly connected products
continued to seek power after compensatory consumption,
indicating no self-repair as a function of the initial compen-
satory task. In contrast, threatened participants who com-
pensated with implicitly connected products did not differ
on power-seeking (M¼5.57, SE ¼.49) compared to non-
threatened participants (M¼5.76, SE ¼.46; F<1), indi-
cating that compensating with implicitly connected
products in the initial compensation task resulted in self-
repair.
In addition, as expected, threatened participants who
were not initially given a chance to compensate (i.e., re-
ceived nonstatus products) did show compensation effects
on the power-seeking measure, with threatened participants
demonstrating marginally more power-seeking (M¼6.02,
SE ¼.49) than nonthreatened participants (M¼4.84,
SE ¼.46; F(1, 124) ¼3.01, p¼.085). There were no sig-
nificant effects for the importance ratings of job character-
istics unrelated to power (all Fs<1).
The findings of experiment 2 provide support for our hy-
pothesis that threatened participants compensate for self-
threats regardless of whether the products are explicitly or
implicitly connected to the threatened self-identity domain.
Threatened participants were willing to pay more for
status-related (but not status-unrelated) products regardless
of their connection to the threatened domain. However,
this compensation was successful only when the connec-
tion between the product and threat domain was implicit.
When the connection was made explicit through the mar-
keting slogan, threatened participants continued to seek
power on a subsequent compensatory opportunity, indicat-
ing that the initial compensatory opportunity did not result
in full repair of the damaged self.
We also addressed the possibility that self-threats in-
crease willingness to pay regardless of whether they are
status-related. Consistent with previous research (Rucker
and Galinsky 2008), threats to power had no effect on will-
ingness to pay for nonstatus products. The general design
and pattern of results provides a very strong test of our hy-
potheses. In particular, the use of the sequential compensa-
tion tasks required the following pattern of results for
hypothesis support: 1) in the first task, compensatory con-
sumption is observed for both explicitly and implicitly con-
nected products, but not for unconnected products; 2) in
FIGURE 3
COMPENSATORY CONSUMPTION AND SELF-REPAIR AS A
FUNCTION OF PRODUCT CONNECTION (EXPERIMENT 2)
6.73
5.57
6.02
4.84
5.76
4.84
1
2
3
4
5
6
7
Explicitly connected Implicitly connected Nonstatus
(unconnected)
Power-seeking
Threat No threat
5.05
6.61
6.16
3.36
5.05
5.75
1
2
3
4
5
6
7
Explicitly connected Implicitl y connected Nonstatus
(unconnected)
Willingness to pay
Threat No threat
Compensatory Consumpon
Self-Repair
10 JOURNAL OF CONSUMER RESEARCH
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
the second task, compensatory consumption is again ob-
served for explicitly connected products, is no longer ob-
served for implicitly connected products, but is now
observed for unconnected products. All of these effects
were in fact demonstrated.
In the next three experiments, we examine the processes
underlying the effects observed in experiments 1 and 2.
Experiments 3a, 3b, and 4 use an experimental causal chain
design to test our hypotheses (Spencer, Zanna, and Fong
2005). The proposed causal chain is that compensating
with explicitly (but not implicitly) connected products trig-
gers rumination, and this rumination in turn impedes self-
repair. Experiment 3a tests the first link in the causal chain
(compensating with explicitly connected products causes
rumination about the self-threat, but compensating with
implicitly connected products does not). Experiment 3b
conceptually replicates experiment 3a, and also varies the
level of explicitness of the product connections to test the
hypothesis that less explicit connections will result in less
rumination about the self-threat than will more explicit
connections. Experiment 4 tests the second link in the
causal chain (rumination about self-threats impedes
self-repair). In addition, we also include manipulation
checks for the product connection manipulations (experi-
ments 3a, 3b).
EXPERIMENT 3A: COMPENSATING
WITH EXPLICITLY CONNECTED
PRODUCTS CAUSES THREAT
RUMINATION
Experiment 3a tested whether the type of connections
(implicit vs. explicit) between the compensatory product
and the threatened identity-domain impacts rumination
about the self-threat. We expected that threatened partici-
pants who compensate with explicitly connected products
would ruminate more about the self-threat than would non-
threatened participants, whereas compensating with im-
plicitly connected products would not increase rumination
compared to nonthreatened participants.
Method
Participants, Design, and Procedure. Participants were
108 Indian business school student volunteers (M
age
¼22.5
years; 37 women). They were randomly assigned to a 2 (in-
telligence threat: yes vs. no) 2 (product connection: ex-
plicit vs. implicit) between-subjects design. First,
participants were threatened or not on the domain of intelli-
gence with the same threat manipulation used in experi-
ment 1. Next, we manipulated product connection and
compensatory consumption with the same choice task used
in experiment 1 (forced compensation) but with a different
associated slogan for the products in the explicitly con-
nected condition. As in experiment 1, in the explicitly
connected product condition, each of the products was pre-
sented along with a slogan that made the products’ connec-
tion with intelligence explicit (e.g., National Geographic
magazine had the slogan “Magazine for the intelligent
reader”), whereas in the implicitly connected product con-
dition the products were presented without slogans.
Participants were then asked to select one of the products
as their choice for the most suitable prize.
Following that, we measured rumination with the same
scale used by Lisjak et al. (2015), which is a three-item
scale adapted from Brunstein and Gollwitzer (1996).
Participants were asked to report the extent to which the
process of considering the three products and making their
choices reminded them of the experience they had recalled
in the writing study (the threat manipulation), gave them
unwanted thoughts about that experience, and made them
think about their weaknesses (1 ¼not at all, 9 ¼very
much; a¼.81). Next, we measured mood related to the
self-threat task by asking participants to indicate how they
felt (uncomfortable, bothered, etc.) when writing about
their past episode (intelligence threat or control) with eight
items along a nine-point scale (1 ¼not at all, 9 ¼very
much; a¼.90). Finally, participants provided demographic
information and were asked about their perceptions of the
purpose of the study.
Manipulation Checks. We conducted a manipulation-
check study for the explicit and implicit connection manip-
ulations separate from the main study to avoid possible
contamination with the threat manipulations (Perdue and
Summers 1986). One hundred members of an online panel
(MTurk) participated in return for a small fee and were ran-
domly assigned to conditions in a one-factor (product con-
nection: explicit vs. implicit) between-subjects design.
Participants were presented with the ad for the
intelligence-related product (National Geographic maga-
zine) that either included the slogan that specifically refer-
enced the self-threat domain of intelligence (explicit
connection) or did not have the slogan (implicit condition;
see web appendix). Participants were then asked to rate the
extent to which the ad made the connection to intelligence
explicit along a nine-point scale (1 ¼not at all explicit,
9¼completely explicit). As expected, participants in the
explicit-connection condition perceived the product to be
more explicitly connected to intelligence (M¼6.30) than
the product in the implicit condition (M¼4.72;
t(98) ¼3.40, p¼.001), indicating that the manipulation
was successful.
Results and Discussion
Exclusion Criteria. Data from five participants (5%)
who did not write anything in the writing task were ex-
cluded, leaving 103 participants for analysis.
RUSTAGI AND SHRUM 11
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
Hypothesis Testing. We expected that for threatened
participants, compensating with explicitly connected prod-
ucts would result in greater rumination about the self-threat
compared to nonthreatened participants, whereas compen-
sating with implicitly connected products would not in-
crease rumination compared to nonthreatened participants.
To test this hypothesis, we conducted a 2 (threat: yes vs.
no) 2 (product connection: explicit vs. implicit)
ANCOVA with rumination as the dependent variable and
mood as the covariate. (The results did not materially
change when the covariate was excluded from the analy-
ses.) The results of this analysis are shown in figure 4. The
main effect of threat was significant (F(1, 98) ¼5.82,
p¼.02). More central to our hypotheses, the threat prod-
uct connection interaction was significant (F(1, 98) ¼3.95,
p¼.05). As expected, threatened participants who com-
pensated with explicitly connected products ruminated
more (M¼5.46, SE ¼.38) than did nonthreatened partici-
pants (M¼3.72, SE ¼.40; F(1, 98) ¼9.24, p¼.003) and
more than did those who compensated with implicitly con-
nected products (M¼4.31, SE ¼.37; F(1, 98) ¼4.86,
p¼.03). The latter two conditions did not differ (F(1,
98) ¼1.12, p¼.29). In addition, nonthreatened participants
reported equally low levels of rumination regardless of
whether they chose products for future studies that were
explicitly connected (M¼3.72, SE ¼.40) or implicitly
connected (M¼4.06, SE ¼.38; F<1) to intelligence.
The results of experiment 3a indicate that compensating
with explicitly connected products causes rumination about
the threat, but compensating with implicitly connected
products does not. The findings support the first part of the
causal chain. In experiment 3b, we conceptually replicate
these findings using a different self-threat domain (power)
and different levels of explicitness of the product connec-
tions with the self-threat domain.
EXPERIMENT 3B: RUMINATION AND
EXPLICITNESS OF PRODUCT
CONNECTIONS
The objectives of experiment 3b were to replicate the
findings of experiment 3a with a different self-threat do-
main (power) and to contrast the effects of different levels
of explicit connections (rather than explicit vs. implicit) on
rumination. To do so, we varied the extent to which the ex-
plicit connection directly referenced the power domain
(power connection) or was semantically related to it (status
connection). Although research suggests that power and
status are linked (Rucker and Galinsky 2008,2009), other
research suggests that they are also different constructs
(Blader and Chen 2012). We have argued that it is the ex-
plicit connection between the compensatory product and
the self-threat domain that induces rumination, because the
explicit connection reactivates the self-threat in memory.
Thus, to the extent that the connection is less explicit, it
should result in less rumination. Consequently, we
expected that the more explicit connection (power-explicit)
would result in more rumination than would the less ex-
plicit (status-explicit) connection. In addition, to the extent
that the status-explicit connection is sufficient to activate
rumination, we expected that threatened participants in
both the power-explicit and status-explicit connection con-
ditions would ruminate more than would those who were
not threatened, implying no self-repair.
Method
Participants, Design, and Procedure. Three hundred
twenty-three members of the MTurk online panel
(M
age
¼38 years; 199 women) participated in return for a
nominal fee and were randomly assigned to conditions in a
2 (power threat: yes vs. no) 2 (product connection:
status-explicit vs. power-explicit) between-subjects design.
Participants were told that they would be participating in
several short studies. In the first study, which used the
same manipulation as in experiment 2, participants were
either threatened or not threatened on power.
Next, we manipulated product connection with the same
choice task design used in experiment 1. Participants in the
status-explicit and power-explicit product conditions were
presented with three products that naturally symbolize sta-
tus: a luxury pen, a luxury watch, and either a silk tie (for
male participants) or a leather handbag (for female partici-
pants). In the status-explicit condition, each of the products
was presented along with a slogan and a brief description
that made the product’s connection with status explicit
FIGURE 4
RUMINATION AS A FUNCTION OF THREAT AND PRODUCT
CONNECTION (EXPERIMENT 3A)
5.46
4.31
3.72
4.06
1
2
3
4
5
6
Explicitly connected Implicitly connected
Ruminaon
Threat No threat
12 JOURNAL OF CONSUMER RESEARCH
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
(e.g., the luxury pen had the slogan “The Wave Pen:
Designed to Impress,” adapted from Rucker, Dubois, and
Galinsky 2011), whereas in the power-explicit condition,
these same products were presented with slogans and prod-
uct descriptions that made the product’s connection to the
domain of power explicit (e.g., the luxury pen had the slo-
gan “The Wave Pen: Designed for Power and Command”).
In each condition, participants were asked to select the
product they preferred the most, followed by a selection of
their second most preferred choice. Following that, we
measured rumination with the same three-item scale used
in experiment 3a (a¼.81) and measured mood with the
same three-item scale used in experiment 1 (a¼.83).
Finally, participants provided demographic information
and were asked about their perceptions of the purpose of
the study.
Manipulation Checks. We conducted a manipulation-
check study for the power-connected and status-connected
manipulations separate from the main study. One hundred
forty-eight members of an online panel (MTurk) partici-
pated in return for a small fee and were randomly assigned
to conditions in a one-factor (product connection: power
explicit vs. status-explicit vs. implicit) between-subjects
design. Participants were presented with the ad for the
power-related product (Wave Pen) that either included the
slogan and brief description that specifically referenced
power (power-explicit connection) or the slogan and brief
description that referenced status (status-explicit condi-
tion). In addition, although we did not have an implicit-
connection condition for experiment 3b, we included an ad
that did not include the slogan but included a brief descrip-
tion unrelated to power or status to facilitate comparisons
across studies (see web appendix). Participants were then
asked to rate the extent to which the ad made the connec-
tion to power explicit along a nine-point scale (1 ¼not at
all explicit, 9 ¼completely explicit). The one-way
ANOVA was significant (F(2, 145) ¼46.14, p<.001). The
product in the power-explicit condition was perceived to
be more explicitly connected to power (M¼7.74) than was
the product in the status-explicit condition (M¼5.67;
t(145) ¼4.88, p<.001), which was perceived to be more
explicitly connected to power than was the product in the
implicit-connection condition (M¼3.67; t(145) ¼4.70,
p<.001). The difference between the power-explicit and
implicit conditions was also significant (t(145) ¼9.61,
p<.001). These results indicate that the manipulation was
successful.
Results and Discussion
Exclusion Criteria. Data from 19 participants (6%)
were excluded from analyses: 10 who did not write any-
thing or wrote non-relevant content in the self-threat task,
and 9 who guessed the purpose of the study. The data were
analyzed for the remaining 304 participants.
Hypothesis Testing. We expected that participants who
compensated with explicitly connected products would ru-
minate more about the power threat than would those who
were not threatened regardless of whether the explicit con-
nection referred to power or status. However, we expected
that threatened participants who compensated with power-
explicit connections would ruminate more than would
those who compensated with status-explicit connections.
To test these hypotheses, we conducted a 2 (power threat:
yes vs. no) 2 (product connection: power-explicit vs.
status-explicit) ANCOVA, with rumination as the depen-
dent variable and mood and gender as covariates. (The
results did not materially differ when the covariates were
excluded from the analyses.)
The results of this analysis can be seen in figure 5. The
main effect of threat was significant (F(1, 298) ¼24.17,
p<.001), and the threat product interaction was margin-
ally significant (F(1, 298) ¼2.73, p¼.099). Gender was
not significant (F<1). Threatened participants who com-
pensated with products explicitly connected to power
(M¼3.20, SE ¼.19) ruminated more than did non-
threatened participants (M¼1.95, SE ¼.18; F(1,
298) ¼21.91, p<.001). Similarly, threatened participants
who compensated with products explicitly connected to
status ruminated more (M¼2.58, SE ¼.20) than did non-
threatened participants (M¼1.95, SE ¼.18; F(1,
298) ¼5.42, p¼.02). More important, as expected, threat-
ened participants who compensated with products explic-
itly connected to power ruminated more (M¼3.20,
SE ¼.19) than did those who compensated with products
explicitly connected to status (M¼2.58, SE ¼.20; F(1,
298) ¼5.25, p¼.02), which indicates that highly explicit
connections induce greater threat rumination than do mod-
erately explicit connections.
The results of experiment 3b indicate that both explicitly
connecting a product to power and explicitly connecting it
to status impede self-repair. These findings suggest that the
concept of status is sufficiently closely linked in memory
to the concept of power that it induces rumination about
the power threat. However, the finding that rumination is
greater when power is explicitly referenced than when sta-
tus is explicitly referenced suggests that the links in mem-
ory are greater in the former condition than in the latter.
Further, although we did not include an implicit-
connection condition in this experiment, the findings from
experiment 3a showing that rumination did not differ be-
tween threat and no-threat conditions for implicit connec-
tions suggests that status-explicit connections induce more
rumination than implicit connections. We interpret this pat-
tern of results as indicating that the less explicit (status)
connection may facilitate some self-repair, but not enough
to fully restore to prethreat levels.
RUSTAGI AND SHRUM 13
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
These findings provide stronger and more nuanced sup-
port for our reasoning that explicit connections reactivate
the previous self-threat by varying the degree to which the
connections are explicit: as explicitness decreases (power
vs. status), so too does rumination about the previous self-
threat. The results of experiment 3b also conceptually rep-
licate experiment 3a, again supporting the first part of the
causal chain. In experiment 4, we test the second part of
the causal chain.
EXPERIMENT 4: RUMINATION IMPEDES
SELF-REPAIR
Experiment 4 tested the second part of the casual chain,
which posits that threat rumination impedes self-repair. To
enhance generalizability, we used a different self-threat do-
main (sociability; Lisjak et al. 2015), a different sample
composition (French students), and a different method of
assessing self-repair. To assess self-repair, we used a
repeated-measures design, with the added benefit of con-
serving power, given that our study was constrained by
class size. We expected that inducing participants to rumi-
nate about the prior sociability threat would impede self-
repair, whereas preventing rumination would facilitate
self-repair.
Method
Participants, Design, and Procedure. Seventy-one
French business school student volunteers (M
age
¼24.1
years; 48 women) were randomly assigned to conditions in
a 2 (time: time 1 vs. time 2) 2 (condition: rumination vs.
distraction) repeated-measures mixed design, with time as
a within-subjects factor and condition as a between-
subjects factor. The data were collected at the beginning of
two different classes at roughly the same time (within the
same week), and participants received extra course credit
for their participation. The study was conducted in two
phases separated by a week in order to reduce the possibil-
ity of demand effects related to repeated measures
(Greenwald 1976). In the first phase (time 1), in a study os-
tensibly about characteristics of the self, all participants
filled out a short self-evaluation survey in which they rated
themselves on their sociability with one item along a nine-
point scale (1 ¼not at all, 9 ¼very much), along with other
decoy measures. This item served as the participants’ base-
line measure of sociability.
A week later (time 2), the same students participated in
another ostensibly unrelated study. First, using the same
cover story as experiment 1, we threatened all participants
on sociability by having them write about a time when they
were not as sociable or friendly toward another person as
they would have liked to be (Lisjak et al. 2015). Next, we
manipulated rumination by inducing some participants to
ruminate about the threat, whereas others were distracted
and thus prevented from ruminating (Bushman et al. 2005;
Lisjak et al. 2015). Participants in the rumination condition
were told to write about what they had done from the start
of the study until the present time and the thoughts and
feelings they had experienced. Participants in the distrac-
tion condition wrote about the physical layout of a building
they knew well (e.g., college campus, supermarket).
Participants then rated themselves on certain personality
characteristics, including sociability, measured with two
items (cordial with others, friendly) along a nine-point
scale (1 ¼not at all, 9 ¼very much). We averaged the two
sociability items to form a composite measure (r¼.59).
Finally, we measured mood with the same three-item
measure (a¼.85) used in experiment 1. To serve as a ma-
nipulation check for rumination, we asked participants to
report the extent to which they ruminated about the experi-
ence they had written about along the same three-item
scale (a¼.84) used in experiments 3a and 3b. Participants
provided demographic information and were asked
whether they had heard about the study before and what
they thought was the purpose of the study.
Results and Discussion
Exclusion Criteria. Data from two participants (3%)
who reported having heard about the experiment before
their participation were excluded from analyses. The data
were analyzed for the remaining 69 participants.
Manipulation Checks. A one-way ANOVA comparing
the effects of condition on the self-reported rumination in-
dex revealed that participants in the rumination condition
FIGURE 5
RUMINATION AS A FUNCTION OF EXPLICIT PRODUCT
CONNECTION (EXPERIMENT 3B).
3.2
2.58
1.95 1.95
1
2
3
4
Power explicit Status explicit
Ruminaon
Threat No threat
14 JOURNAL OF CONSUMER RESEARCH
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
ruminated more about the threat (M¼5.54, SD ¼1.88)
than did those in the distraction condition (M¼2.16,
SD ¼1.84; F(1, 67) ¼56.99, p<.001), indicating that the
rumination manipulation was successful.
Hypothesis Testing. We expected that rumination
would decrease sociability ratings relative to baseline rat-
ings, but expected no such lowering of sociability ratings
for those who were distracted from ruminating. To test
these hypotheses, we conducted a 2 (time: time 1 vs. time
2) 2 (condition: rumination vs. distraction) repeated-
measures ANCOVA with time as a within-subjects factor,
condition as a between-subjects factor, mood as a covari-
ate, and sociability rating as the dependent variable. (The
results did not materially differ when the covariate was ex-
cluded from the analyses.) The expected time condition
interaction was significant (F(1, 66) ¼4.63, p¼.04). The
results of this analysis can be seen in figure 6. Participants
who were induced to ruminate about the prior self-threat
rated themselves lower on sociability at time 2 (M¼6.97,
SE ¼.18) compared to their baseline self-rating at time 1
(M¼7.31, SE ¼.21; diff. ¼.34; F(1, 66) ¼4.88, p¼.03).
In contrast, participants who were prevented (distracted)
from ruminating about the same self-threat did not differ
on their sociability ratings between time 1 (M¼7.10,
SE ¼.20) and time 2 (M¼7.22, SE ¼.17; diff. ¼.12;
F<1), and the difference between time 1 and time 2 rat-
ings was greater in rumination conditions (M
diff
¼.34) than
in distraction conditions (M
diff
¼.12), as evidenced by
the significant interaction.
Experiment 4 supports our hypothesis that ruminating
about a threat undermines self-repair. After experiencing a
self-threat in the sociability domain, participants who were
induced to ruminate about the self-threat rated themselves
as less sociable compared to their baseline measure of so-
ciability obtained one week earlier. In contrast, sociability
ratings of those who did not ruminate (distraction condi-
tion) did not differ from their baseline measure. These
findings provide evidence that rumination impedes self-
repair, supporting the second part of the causal chain.
Together, the results from experiments 3a, 3b, and 4 indi-
cate that compensating with explicitly connected products
increases rumination about the self-threat, and this rumina-
tion in turn impedes self-repair. In contrast, compensating
with implicitly connected products does not cause rumina-
tion, and consequently, self-repair is successful.
Thus far, we have provided evidence that compensating
with explicitly connected products in response to self-
threats impedes self-repair, but compensating with
implicitly connected products facilitates self-repair.
In experiment 5, we explore a boundary condition to these
effects by demonstrating conditions under which explicitly
connected products can also facilitate self-repair.
EXPERIMENT 5: EXPLICIT PRODUCT
CONNECTIONS CAN FACILITATE
REPAIR IF THE SELF-THREAT IS
IMPLICIT
Experiment 5 tested conditions under which even explic-
itly connected products may facilitate self-repair to shed
additional light on the underlying mechanisms. More spe-
cifically, we tested the proposition that when the self-threat
itself is implicit, self-repair will occur for both explicitly
and implicitly connected products.
Method
Participants, Design, and Procedure. Seventy-nine
Indian undergraduates (M
age
¼19.1 years; 43 women) par-
ticipated in return for the possibility of winning a lottery
for an Amazon gift voucher of Rs. 500–1500 (approxi-
mately US$8–$24). Participants were told that they would
be participating in several short studies and were randomly
assigned to conditions in a one-factor (intelligence threat:
explicit vs. implicit vs. none) between-subjects design.
First, participants were either threatened on the domain
of intelligence with explicit or implicit manipulations, or
were not threatened. Those in the explicit threat condition
were threatened on the domain of intelligence with the
same manipulation used in experiment 1. Participants in
the implicit threat condition were threatened on intelli-
gence using a writing task adapted from Gao et. al (2009,
study 3). Unlike those in the recall task in the explicit
FIGURE 6
SOCIABILITY AS A FUNCTION OF RUMINATION CONDITION
(EXPERIMENT 4).
7.31
7.1
6.97
7.22
6.2
6.4
6.6
6.8
7
7.2
7.4
Ruminaon Distracon
Sociability
Time 1 Time 2
RUSTAGI AND SHRUM 15
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
threat condition, the writing themes in the implicit threat
condition have no overt connection to failures on the do-
main of intelligence. Specifically, participants were asked
to share two experiences each on two different themes pre-
sented in sequence. In step 1, participants were asked to
list two life experiences in which they felt a great deal of
doubt or uncertainty. In the next step, they were asked to
list two experiences that made them feel intellectually
smart and bright. Based on past research (Gao et al. 2009;
Petty, Bri~
nol, and Tormala 2002), we expected that partici-
pants would attribute their general self-doubt induced by
sharing the first two experiences in step 1 to having doubts
on the domain of intelligence about which they wrote sub-
sequently. Participants in the no-threat condition also wrote
two experiences each on two different themes. However,
for the first theme, instead of listing experiences in which
they had doubt about their intelligence, they began by list-
ing two experiences in which they felt confident and cer-
tain about themselves and then shared two experiences that
made them feel intellectually smart and bright.
Next, we provided all participants an opportunity to
compensate with products explicitly connected to the do-
main of intelligence using the same compensation task and
products used in experiment 1. After completing a filler
task that asked them about their lifestyle preferences, par-
ticipants completed an ostensibly new study that assessed
how long they were willing to wait in line under two differ-
ent scenarios (adapted from Inesi et al. 2011). In the first,
threat-related scenario, participants were asked to imagine
that a “Knowledge Exhibition” was being held to showcase
the work and contributions of leading scientists, artists, and
writers, but there was a long line at the venue to gain entry
to the exhibition. At the end of the scenario, participants
were asked to indicate along a nine-point scale (1 ¼less
than 5 minutes, 9 ¼more than 40 minutes) how long they
would be willing to wait in line in order to attend the exhi-
bition. In the second, threat-unrelated scenario, participants
were asked to imagine that a new convenience store had re-
cently opened near their home and that they had reached
the store at its opening hour just to try it out. However, the
store had not yet opened. They were asked to indicate
along a nine-point scale (1 ¼less than 2 minutes, 9 ¼more
than 16 minutes) how long they were willing to wait for the
store to open. Finally, mood was measured with the three-
item scale used in study 1 (a¼.83), and participants pro-
vided demographic information and were asked about their
perceptions of the purpose of the study.
Web Appendix Experiment 1. To ensure that the im-
plicit manipulation produced the expected effects, we con-
ducted a separate experiment. Although the implicit threat
manipulation we employed has been used successfully in
previous studies (Gao et al. 2009;Petty et al. 2002), partic-
ipant characteristics and experimental methods in our study
differed from previous ones, and demonstration of
spontaneous compensatory consumption following the im-
plicit threat is critical to the interpretation of experiment 5.
In this experiment, we used the same implicit threat manip-
ulation (intelligence threat) and participants drawn from
the same general participant pool as experiment 5, and also
included an explicit threat manipulation that was used in
experiment 1 (full details can be found in the web appen-
dix). The results showed that the implicit and explicit
threat manipulations were successful. Both participants
who were explicitly threatened and who were implicitly
threatened on intelligence expressed a greater preference
for an intelligence-related product than did those who were
not threatened, and product preference did not differ be-
tween implicit and explicit threat conditions.
Results and Discussion
Exclusion Criteria. Data from three participants (4%)
who did not write anything in the writing tasks were ex-
cluded from analyses, leaving 76 participants for the
analyses.
Hypothesis Testing. We expected that participants in
the explicit threat condition would continue to compensate
for the intelligence threat, and thus would be willing to
wait longer in the line for attending the intelligence sym-
bolic knowledge exhibition than would those in the no-
threat conditions, indicating no self-repair. In contrast, we
expected that those in the implicit threat condition would
not continue to compensate, and thus would show no dif-
ferences in willingness to wait between the implicit threat
and no-threat conditions, indicating self-repair. In addition,
we did not expect willingness to wait for the intelligence-
unrelated convenience store to differ across the threat
conditions.
To test these hypotheses, we conducted a one-way
ANCOVA with threat condition as the independent vari-
able, mood as a covariate, and willingness to wait as the
dependent variable. (The results did not materially change
when the covariate was excluded from the analyses.) The
results of this analysis can be seen in figure 7. For the
threat-related scenario (Knowledge Exhibition), the main
effect of threat condition was significant (F(2, 72) ¼3.54,
p¼.034). Participants in the explicit threat condition were
willing to wait longer (M¼5.24, SE ¼.43) than were par-
ticipants in the implicit threat condition (M¼3.69,
SE ¼.43; p¼.014) and the no-threat condition (M¼4.03,
SE ¼.43; p¼.05), indicating no self-repair for the explicit
threat condition, consistent with previous experiments. In
contrast, there was no difference in willingness to wait be-
tween the implicit threat and no-threat conditions (p¼.58),
indicating self-repair. Further, as expected, there was no ef-
fect of threat condition on willingness to wait in the threat-
unrelated scenario (convenience store; F<1).
16 JOURNAL OF CONSUMER RESEARCH
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
The results of experiment 5 provide an important bound-
ary condition for the findings of our previous experiments
by showing that an explicit connection between the com-
pensatory product and the self-threat domain can result in
self-repair, under certain conditions. Specifically, when
self-concept deficits arise from implicit self-threats—ones
in which there is no specific self-threat, but instead a more
generalized reduction in self-confidence—explicitly con-
nected products can also lead to self-repair. However, as in
previous experiments, compensating with explicitly con-
nected products impedes self-repair when the self-threat is
explicit. These results suggest that, consistent with the
findings of experiment 3a and 3b, and web appendix exper-
iment 1, when a threat has explicit linkages to a failure on
the self-concept domain, explicit product connections reac-
tivate the previous self-threat, which induces rumination
and thwarts self-repair. However, this process occurs only
when the self-threat is explicitly stored in memory. When
the self-threat is implicit, there is no explicit self-threat to
reactivate, and thus a compensatory product’s explicit con-
nection to a self-threat domain should not cause rumina-
tion, and thus not impede self-repair.
GENERAL DISCUSSION
People generally want to feel good about themselves.
However, occasionally, or maybe even often, people en-
counter threats to important aspects of their self-concept.
Numerous studies demonstrated that one way that people
try to repair their damaged self-concept is through sym-
bolic consumption (Rucker and Galinsky 2013): when an
aspect of their self-concept is threatened, people seek out
products that signal success on the threatened self-domain
(within-domain compensatory consumption). Although this
within-domain compensatory effect has been observed
across many self-relevant domains, surprisingly little re-
search has addressed the efficacy of compensatory con-
sumption, and the few studies that have done so have
produced mixed results.
In the present research, we attempted to reconcile these
disparate findings. Across seven experiments with diverse
participants using multiple measures and manipulations,
we showed that within-domain compensatory consumption
can restore the damaged self, but only under certain condi-
tions. When the association between the symbolic product
and the self-threat domain is not made explicit, symbolic
consumption can successfully restore the damaged self-
concept. However, in many cases, the connection between
the product and a particular self-concept domain may be
made explicit through a product name, marketing slogan,
or advertising tagline. Such explicit connections are gener-
ally viewed as good marketing: brands want to convey
their positioning on particular attributes or provide clear in-
formation about what a product can offer.
Our research shows that such explicit connections can
actually have a detrimental effect on restoring a damaged
self-concept. We show that when inherent (implicit) con-
nections between the product and the damaged self-
concept that would otherwise result in a successful self-
concept repair are made explicit, self-repair is thwarted.
The explicit connection serves to reactivate the threat in
memory, which causes individuals to ruminate about
their insecurities and shortcomings in that domain
(Lisjak et al. 2015), and this rumination impedes self-
repair. However, implicit connections do not trigger such
ruminative thoughts, allowing self-repair to occur. Thus,
compensating with explicitly connected products has the
insidious effect of actually prolonging rather than repairing
the damaged self-concept. Unfortunately for consumers,
our research also shows that they are apparently unaware
of these differences in the efficacy of compensatory
FIGURE 7
WILLINGNESS TO WAIT AS A FUNCTION OF THREAT TYPE
(EXPERIMENT 5).
Convenience store
4.84
4.01 4.26
1
2
3
4
5
6
Explicit threat Implicit threat No threat
Willingness to wait in queue
Knowledge exhibion
5.24
3.69
4.03
1
2
3
4
5
6
Explicit threat Implicit threat No threat
Willingness to wait in queue
RUSTAGI AND SHRUM 17
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
consumption via implicitly versus explicitly connected
products, as they show equal levels of compensatory con-
sumption regardless of the explicitness of the product con-
nection to the damaged self-concept domain.
Contributions
Our research makes several contributions. First, it poten-
tially reconciles seemingly conflicting findings in previous
research. Some research has demonstrated that within-
domain compensatory consumption can successfully repair
a damaged self-identity (Gao et al. 2009), whereas other re-
search has shown that within-domain compensation actu-
ally impedes self-repair (Lisjak et al. 2015). Our research
can account for these different findings. When the sym-
bolic connections between the self-threat domain and the
compensatory product are made explicit, as was the case in
the studies reported by Lisjak et al., the explicit connec-
tions induce rumination about the self-threat, which
impairs self-control and impedes self-repair. However,
when the connections between the self-threat and compen-
satory product are implicit (inherently symbolic of the self-
threat domain), and no explicit connections are made, as
was the case in the research reported by Gao et al. (2009),
compensatory consumption can successfully repair the
damaged self-concept.
Although our research fully replicates the findings of
Lisjak et al. (2015) that self-repair is impeded when the
connections between the compensatory product and the
self-threat domain are made explicit, our research also sug-
gests a slightly different interpretation of their findings.
They proposed that within-domain compensatory con-
sumption caused rumination about the self-threat, and
thereby reduced self-control. However, our research sug-
gests that it is not within-domain compensation per se that
triggers rumination, but rather the explicit connections be-
tween the products and the threatened self-domain.
A second contribution pertains to the processes underly-
ing compensatory consumption more generally, and how
the self-repair process does (or does not) work. We have
argued that the explicit connections reactivate the self-
threat, which induces rumination, and the extent to which
this reactivation occurs is related to the explicitness of the
connection and the strength of the association in memory
between the explicit connection and the self-threat domain.
Consequently, as we showed in experiment 3b, less explicit
connections (status connection following a power threat)
result in less rumination than do more explicit connections
(power connection following a power threat), but both re-
sult in more rumination than implicit connections.
Thus, the fundamental issue is not about explicit or im-
plicit connections per se, but rather the extent to which
aspects of the product reactivate the self-threat and induce
rumination. One implication is that there may be instances
in which even implicit connections induce rumination and
impede self-repair. For some individuals, the inherent sym-
bolic (implicit) associations may be so strongly linked to a
self-concept domain that the product itself—sans any ex-
plicit connection—may reactivate the self-threat. For ex-
ample, highly materialistic consumers tend to be overly
concerned with status and believe that possessions are indi-
cators of success, and they also tend to have chronic self-
deficits in the identity domain of power (Richins and
Dawson 1992). Similarly, intelligence is likely one of the
more central self-concept domains for academics. Thus,
products that are inherently symbolic of these self-concept
domains (luxury cars, prestigious journal publications)
may automatically activate a self-concept domain for those
who have chronically strong links between the self-identity
symbolism of the product and chronic deficits on the self-
concept domain. For such individuals, compensating with
implicitly connected products may also impede self-repair.
Our research also has implications for consumer well-
being. We show that although compensating for self-
threats via implicitly connected products is more effective
in restoring the self-concept than compensating with ex-
plicitly connected products, consumers are not well cali-
brated in terms of the relative effectiveness of the two
compensatory modes. Thus, threatened consumers may
end up wastefully and impulsively spending on explicitly
connected products even when it does not provide any self-
repair benefits.
Finally, our research has implications for materialism re-
search. Although materialism is typically operationalized
as stable traits and values (Belk 1985;Richins and Dawson
1992), to the extent that products are used to bolster
aspects of self-identity, their consumption arguably repre-
sents materialistic behaviors (Rustagi and Shrum 2018;
Shrum et al. 2013). Most research on materialism has fo-
cused on the negative relation between materialism and
well-being (Dittmar et al. 2014). However, our research
suggests that materialistic behavior, at least in the form of
compensatory consumption, can have positive utility under
some conditions.
Alternative Explanations and Future Research
We have argued that contextual features such as explicit
slogans or product names can reactivate and cue thoughts
about a previous self-threat, which causes consumers to
continue to try to bolster or repair their damaged self-
concept. A slightly different explanation is that the explicit
cues actually increase the magnitude of the threat (make it
stronger), which induces rumination and impedes self-
repair. Although this explanation is plausible, our data sug-
gest that this is not the case. In particular, if the explicit
connections in effect increase the magnitude of the manip-
ulation, then we would expect that the initial compensatory
consumption effects would be greater for explicitly con-
nected products than for implicitly connected products.
18 JOURNAL OF CONSUMER RESEARCH
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
However, as the findings of experiment 2 indicate, the ini-
tial compensatory effects did not differ as a function of
connection conditions.
Another alternative explanation for the effects of explicit
connections is that the explicit cues may induce worry
about what others will think about their motives for the
purchase and thus how it may potentially reflect on them-
selves, and this worry drives the rumination effects. This
explanation is also plausible, and consistent with findings
showing that receiving social validation in a self-domain
that has been threatened provides a sense of goal fulfill-
ment, which can facilitate self-repair (Brunstein and
Gollwitzer 1996;Lisjak et al. 2015, experiment 4).
However, this alternative has a difficult time explaining
the pattern of results we observed. First, if the explicit cue
induces consumers to worry that others may think poorly
of them because of their choice, then it is unclear why con-
sumers would show a greater preference or would be will-
ing to pay more for these explicitly connected products in
the first place. Second, this alternative explanation also has
difficulty accounting for the differences noted between
power-connected and status-connected cues (experiment
3b). More specifically, we have argued that the differences
are the result of different levels of explicitness, which re-
sult in less likelihood or less strength of reactivation of the
self-threat because the connection between the cue and the
self-threat domain is less explicit. However, the more ex-
plicit (power-connected) and less explicit (status-con-
nected) connections should not produce different levels of
worry about what others think of the choice.
Although we have argued that our research provides a
compelling reconciliation of previous research on compen-
satory consumption and self-repair, there is at least one
unpublished study that has reported successful self-repair
with compensatory products that are explicitly connected
to the self-threat domain. Rucker et al. (2011) (reported in
Rucker, Galinsky, and Dubois 2012) found that compensat-
ing for a power threat with a pen advertised as a status pen
restored participants’ feelings of power, results that are
clearly contrary to our findings and theoretical reasoning.
Although our research cannot directly resolve the apparent
discrepancy, there are important differences between that
study and ours that may explain the discrepancies, and thus
potentially represent avenues for future research. One dif-
ference pertains to the compensatory process. In our stud-
ies, participants were asked to choose products that were
connected to the self-threat domain, whereas participants
in Rucker et al. (2011) were given a product (status pen) to
evaluate and were allowed to keep the product as a gift.
Thus, it may be that physical (as opposed to mental) con-
sumption has a greater restorative effect. A second differ-
ence also pertains to the compensatory process. In our
studies, participants made choices among alternatives
(three explicitly connected products), whereas no choice
process occurred in the Rucker et al. study. The process of
deciding which compensatory product to choose or how
much to pay for it requires a certain amount of consider-
ation and even self-reflection, which may facilitate or in-
crease the probability that the prior self-threat may be
reactivated, thus spurring rumination. In contrast, receiving
a product as a gift requires no such mental processes, and
thus reactivation and rumination may be avoided.
Although we have shown that compensatory consump-
tion can successfully restore a damaged self-concept, and
thus may be useful for consumers in some contexts, there
are important limitations that we believe are worth noting.
First, for obvious ethical reasons, the laboratory manipula-
tions of self-threats (e.g., recalling a time when an aspect
of the self was threatened) are relatively benign, and most
research shows that such manipulations are not particularly
upsetting and dissipate quickly. However, in daily life,
these threats may be remarkably traumatizing for some,
and also may occur frequently. For example, a person may
not merely receive a demotion or have his or her job re-
sponsibilities curtailed, but may actually be fired. In such
cases, it is unlikely that a single instance of compensatory
consumption will successfully repair the damaged self-
concept. Clearly, researchers should not subject partici-
pants to traumatic self-threats, but other methods (e.g., ex-
perience sampling) may be useful in documenting
compensatory consumption processes and whether they are
successful over time.
A related concern pertains to the scope of compensatory
effects in everyday life, and the extent to which self-threats
are mostly implicit or explicit. That is, our research shows
that compensating for implicit threats restores the self-
concept regardless of the explicitness of the product con-
nection, and thus if most threats are implicit, little harm
may result from compensatory consumption. However, as
we noted, many self-threats are clear: failing an exam, be-
ing denied membership in a club, losing an election.
Moreover, individuals who are chronically lacking in self-
worth may perceive even seemingly small slights to be
highly threatening. In such cases, we argue that the self-
threat is explicit and likely to be highly accessible, perhaps
for a lengthy amount of time.
Similarly, in many instances, the form of compensatory
consumption can be relatively benign. One particularly in-
teresting aspect of research on compensatory consumption
is that the “consumption” part need not involve purchase
or use. Merely choosing threat-related products for
researchers to use in future studies (Gao et al. 2009, this
study), being entered into a lottery for threat-related prod-
ucts (Gao et al. 2009), receiving a free gift that was adver-
tised as threat-related (Rucker et al. 2011), or expressing
willingness to pay for threat-related products (this study) is
sufficient to repair the damaged self. Moreover, in most
instances, the threat-related products were not at all expen-
sive. Thus, there are likely few adverse consequences of
imagining owning an Aston Martin, buying a magazine
RUSTAGI AND SHRUM 19
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
such as the Economist, or going to a museum, to compen-
sate for self-threats. However, not all modes of compensa-
tory consumption may be so benign. Spending on products
to remedy a self-threat can just as easily involve expensive
products (e.g., luxury goods) that may stretch or break
budgets, potentially increasing debt and directing resources
away from needed goods.
DATA COLLECTION INFORMATION
The data for experiment 1 were collected by the first au-
thor from Amazon’s MTurk online panel on August 2 and
3, 2016. Data for experiment 2 were collected by the first
author from Amazon’s MTurk online panel from July 21 to
August 3, 2015. Data collection for experiment 3a was co-
ordinated by the first author with staff members of Fortune
Institute of International Business, New Delhi, India, who
supervised the data collection through an on-campus pa-
per-and-pencil-based study conducted with MBA students
from October 21 to 23, 2015. Data for experiment 3b were
collected by the first author from Amazon’s MTurk online
panel in April 2017. Data for experiment 4 were jointly
collected by both authors through a paper-and-pencil on-
campus study conducted at HEC Paris. The participants
were graduate students of HEC Paris. The data were col-
lected in two phases from October 12 to 21, 2015. Data for
experiment 5 were collected by the first author through an
on-campus paper-and-pencil-based study conducted with
Indian undergraduate students in April 2017. The first au-
thor analyzed all data under the supervision of the second
author.
REFERENCES
Belk, Russell W. (1985), “Materialism: Trait Aspects of Living in
the Material World,” Journal of Consumer Research, 12 (3),
265–80.
Bhattacharjee, Amit, Jonah Berger, and Geeta Menon (2014),
“When Identity Marketing Backfires: Consumer Agency in
Identity Expression,” Journal of Consumer Research, 41 (2),
294–309.
Blader, Steven L. and Ya-Ru Chen (2012), “Differentiating
Effects of Power and Status: A Justice Perspective,” Journal
of Personality and Social Psychology, 102 (5), 994–1014.
Brunstein, Joachim C. and Peter M. Gollwitzer (1996), “Effects of
Failure on Subsequent Performance: The Importance of Self-
Defining Goals,” Journal of Personality and Social
Psychology, 70 (2), 395–407.
Bushman, Brad J., Angelica M. Bonacci, William C. Pedersen,
Eduardo A. Vasquez, and Norman Miller (2005), “Chewing
on It Can Chew You Up: Effects of Rumination on Triggered
Displaced Aggression,” Journal of Personality and Social
Psychology, 88 (6), 969–83.
Chernev, Alexander, Ryan Hamilton, and David Gal (2011),
“Competing for Consumer Identity: Limits to Self-
Expression and the Perils of Lifestyle Branding,” Journal of
Marketing, 75 (3), 66–82.
Chusmir, Leonard H. and Barbara Parker (1984), “Dimensions of
Need for Power: Personalized vs. Socialized Power in
Female and Male Managers,” Sex Roles, 11 (9–10), 759–69.
Cutright, Keisha M., Eugenia C. Wu, Jillian C. Banfield, Aaron C.
Kay, and Gavan J. Fitzsimons (2011), “When Your World
Must Be Defended: Choosing Products to Justify the
System,” Journal of Consumer Research, 38 (1), 62–77.
Dittmar, Helga, Rod Bond, Megan Hurst, and Tim Kasser (2014),
“The Relationship between Materialism and Personal Well-
Being: A Meta-Analysis,” Journal of Personality & Social
Psychology, 107 (5), 879–924.
Duclos, Rod, Echo Wen Wan, and Yuwei Jiang (2013), “Show Me
the Honey! Effects of Social Exclusion on Financial Risk-
Taking,” Journal of Consumer Research, 40 (1), 122–35.
Gao, Leilei, S. Christian Wheeler, and Baba Shiv (2009), “The ‘Shaken
Self’: Product Choices as a Means of Restoring Self-View
Confidence,” Journal of Consumer Research, 36 (1), 29–38.
Greenwald, Anthony G. (1976), “Within-Subjects Designs: To
Use or Not to Use?” Psychological Bulletin, 83 (2), 314–20.
Higgins, E. Tory (1987), “Self-Discrepancy: A Theory Relating
Self and Affect,” Psychological Review, 94 (3), 319–40.
Inesi, M. Ena, Simona Botti, David Dubois, Derek D. Rucker, and
Adam D. Galinsky (2011), “Power and Choice: Their
Dynamic Interplay in Quenching the Thirst for Personal
Control,” Psychological Science, 22 (8), 1042–8.
Kim, Soo and David Gal (2014), “From Compensatory
Consumption to Adaptive Consumption: The Role of Self-
Acceptance in Resolving Self-Deficits,” Journal of
Consumer Research, 41 (2), 526–42.
Kim, Soo and Derek D. Rucker (2012), “Bracing for the
Psychological Storm: Proactive versus Reactive
Compensatory Consumption,” Journal of Consumer
Research, 39 (4), 815–30.
Leary, Mark R., Ellen S. Tambor, Sonja K. Terdal, and Deborah
L. Downs (1995), “Self-Esteem as an Interpersonal Monitor:
The Sociometer Hypothesis,” Journal of Personality and
Social Psychology, 68 (3), 518–30.
Lee, Jaehoon and L. J. Shrum (2012), “Conspicuous Consumption
versus Charitable Behavior in Response to Social Exclusion:
A Differential Needs Explanation,” Journal of Consumer
Research, 39 (3), 530–44.
Lee, Jaehoon, L. J Shrum, and Youjae Yi (2017), “The Role of
Communication Norms in Social Exclusion Effects,” Journal
of Consumer Psychology, 27 (1), 108–16.
Levav, Jonathan and Rui (Juliet) Zhu (2009), “Seeking Freedom
through Variety,” Journal of Consumer Research, 36 (4),
600–10.
Lisjak, Monika, Andrea Bonezzi, Soo Kim, and Derek D. Rucker
(2015), “Perils of Compensatory Consumption: Within-Domain
Compensation Undermines Subsequent Self-Regulation,”
Journal of Consumer Research, 41 (5), 1186–203.
Mandel, Naomi, Derek D. Rucker, Jonathan Levav, and Adam D.
Galinsky (2017), “The Compensatory Consumer Behavior
Model: How Self-Discrepancies Drive Consumer Behavior,”
Journal of Consumer Psychology, 27 (1), 133–46.
Martin, Leonard L. (1986), “Set/Reset: Use and Disuse of
Concepts in Impression Formation,” Journal of Personality
and Social Psychology, 51 (3), 493–504.
Mead, Nicole L., Roy F. Baumeister, Tyler F. Stillman, Catherine
D. Rawn, and Kathleen D. Vohs (2011), “Social Exclusion
Causes People to Spend and Consume Strategically in the
Service of Affiliation,” Journal of Consumer Research,37
(5), 902–19.
20 JOURNAL OF CONSUMER RESEARCH
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018
Perdue, Barbara C. and John O. Summers (1986), “Checking the
Success of Manipulations in Marketing Experiments,”
Journal of Marketing Research, 23 (4), 317–26.
Petty, Richard E., Pablo Bri~
nol, and Zakary L. Tormala (2002),
“Thought Confidence as a Determinant of Persuasion: The
Self-Validation Hypothesis,” Journal of Personality and
Social Psychology, 82 (5), 722–41.
Richins, Marsha L. and Scott Dawson (1992), “A Consumer
Values Orientation for Materialism and Its Measurement:
Scale Development and Validation,” Journal of Consumer
Research, 19 (3), 303–16.
Rucker, Derek D., David Dubois, and Adam D. Galinsky (2011),
“Status as a Salve for a Loss of Power,” unpublished manu-
script, Northwestern University, Evanston, IL 60208.
Rucker, Derek D. and Adam D. Galinsky (2008), “Desire to
Acquire: Powerlessness and Compensatory Consumption,”
Journal of Consumer Research, 35 (2), 257–67.
———. (2009), “Conspicuous Consumption versus Utilitarian
Ideals: How Different Levels of Power Shape Consumer
Behavior,” Journal of Experimental Social Psychology,45
(3), 549–55.
———. (2013), “Compensatory Consumption,” in The Routledge
Companion to Identity and Consumption, ed. Ayalla A.
Ruvio and Russell W. Belk, New York: Routledge, 207–15.
Rucker, Derek D., Adam D. Galinsky, and David Dubois (2012),
“Power and Consumer Behavior: How Power Shapes Who
and What Consumers Value,” Journal of Consumer
Psychology, 22 (3), 352–68.
Rustagi, Nimish and L. J. Shrum (2018), “Materialism:
Conceptualizations, Antecedents, and Consequences,” in The
Routledge Companion to Consumer Behavior, ed. Michael R.
Solomon and Tina M. Lowrey, New York: Routledge, 21–37.
Shrum, L. J., Nancy Wong, Farrah Arif, Sunaina K. Chugani,
Alexander Gunz, Tina M. Lowrey, Agnes Nairn, Mario
Pandelaere et al. (2013), “Reconceptualizing Materialism as
Identity Goal Pursuits: Functions, Processes, and
Consequences,” Journal of Business Research, 66 (8),
1179–85.
Smith, Pamela K., Dani€
el H. J. Wigboldus, and Ap Dijksterhuis
(2008), “Abstract Thinking Increases One’s Sense of Power,”
Journal of Experimental Social Psychology, 44 (2), 378–85.
Spencer, Steven J., Mark P. Zanna, and Geoffrey T. Fong (2005),
“Establishing a Causal Chain: Why Experiments Are Often
More Effective than Mediational Analyses in Examining
Psychological Processes,” Journal of Personality and Social
Psychology, 89 (6), 845–51.
Vohs, Kathleen D. and Todd F. Heatherton (2004), “Ego Threat
Elicits Different Social Comparison Processes among High
and Low Self-Esteem People: Implications for Interpersonal
Perceptions,” Social Cognition, 22 (1), 168–91.
Wicklund, Robert A. and Peter M. Gollwitzer (1982), Symbolic
Self-Completion, Hillsdale, NJ: Lawrence Erlbaum
Associates.
Williams, Kipling D. (2007), “Ostracism,” Annual Review of
Psychology, 58, 425–52.
Wyer, Robert S. and Thomas K. Srull (1989), Memory and
Cognition in Its Social Context, Hillsdale, NJ: Erlbaum.
RUSTAGI AND SHRUM 21
Downloaded from https://academic.oup.com/jcr/advance-article-abstract/doi/10.1093/jcr/ucy064/5056709 by Hec Paris user on 06 December 2018