Putting Escalation of Commitment in Context: A Multi-level Review and Analysis

Abstract

Organizations, and the groups and individuals within them, sometimes face the thorny dilemma of whether or not to continue failing courses of action. Escalation of commitment describes the tendency to “carry on” with such questionable endeavors, regardless of whether doing so is likely to result in success. Despite the wide variability between and within organizations, early research on escalation of commitment largely ignored the role of context. Since then, an impressive volume of studies on the context of escalation has accumulated, and we present a fresh take on the literature by reviewing and analyzing this work. Using a multilevel framework, we elucidate how escalation is influenced by factors at the group and organizational levels, as well as by factors external to the organization. Ultimately, we aim to provide a more comprehensive and nuanced understanding of escalation to stimulate new avenues of research and offer insights into how decision-makers can respond more effectively to escalation dilemmas.

Full text

rAcademy of Management Annals
2018, Vol. 12, No. 1, 178–207.
https://doi.org/10.5465/annals.2016.0046
PUTTING ESCALATION OF COMMITMENT IN CONTEXT: A
MULTILEVEL REVIEW AND ANALYSIS
DUSTIN J. SLEESMAN
1
University of Delaware
ANNA C. LENNARD
Michigan State University
GERRY MCNAMARA
Michigan State University
DONALD E. CONLON
Michigan State University
Organizations, and the groups and individuals within them, sometimes face the thorny
dilemma of whether or not to continue failing courses of action. Escalation of commit-
ment describes the tendency to “carry on”with such questionable endeavors, regardless
of whether doing so is likely to result in success. Despite the wide variability between
and within organizations, early research on escalation of commitment largely ignored
the role of context. Since then, an impressive volume of studies on the context of esca-
lation has accumulated, and we present a fresh take on the literature by reviewing and
analyzing this work. Using a multilevel framework, we elucidate how escalation is
influenced by factors at the group and organizational levels, as well as by factors ex-
ternal to the organization. Ultimately, we aim to provide a more comprehensive and
nuanced understanding of escalation to stimulate new avenues of research and offer
insights into how decision-makers can respond more effectively to escalation dilemmas.
INTRODUCTION
After being asked for his view on the most impor-
tant secret of success, Mark Zuckerberg, the co-
founder and CEO of Facebook, responded with
a simple piece of advice: “Don’t give up”(Bort, 2015).
Such counsel is hardly unique to this self-made
billionaire or his organization. The belief that people
should persist in the face of difficulty is widely
cherished throughout a number of important situa-
tions in our lives, including decisions about business
(e.g., continuing a failing venture), government
(e.g., allocating more funds to an unsuccessful pro-
gram), and personal relationships (e.g., remaining
in a troubled marriage).
However, this folk wisdom has been challenged by
a large body of work on escalation of commitment—
the act of “carrying on”with questionable or fail-
ing courses of action. Specifically, after investing
significant resources (such as time, money, or ef-
fort) in pursuit of a goal and receiving negative
feedback about the investment, decision-makers
typically maintain or increase commitment to
their goal, despite considerable uncertainty about
whether this will result in success (Brockner,
1992; Staw, 1976; Staw & Ross, 1987). Hundreds
of academic articles on escalation of commitment
have been published in business and related fields,
such as economics and social psychology (Sleesman,
Conlon, McNamara, & Miles, 2012).
After Staw (1976) introduced the escalation of
commitment phenomenon, most initial work on the
topic focused on psychological explanations and
was primarily concerned with the individual level of
analysis. In a recent reflection on these adolescent
years of the escalation literature, Staw (2016) noted
that early on, he had felt “uncomfortable with the
near total reliance on laboratory research to test
the phenomenon”(p. 10). His concern that escala-
tion research was constrained to decontextualized,
highly controlled environments was shared by
Bowen (1987), who urged researchers that “careful
We are grateful for the support and guidance provided
by Madan Pillutla and the review team.
1Corresponding author.
178
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s express
written permission. Users may print, download, or email articles for individual use only.

attention must be paid to the context of the decision
under study”(p. 63). Similarly, Brockner (1992)
called for more contextualized work beyond the in-
dividual level of analysis, and Staw (1997) later
echoed this need, asserting that “the completion of
additional field studies, both quantitative and qual-
itative in nature, should be given our highest prior-
ity”(pp. 211–212). In sum, the early years of the
escalation literature were marked by a lack of atten-
tion to the context of escalation, resulting in a poor
understanding of the rich and complex dynamics
underlying escalation behavior in organizations.
Researchers subsequently began to fill this gap, and
a number of studies on the contextual aspects of esca-
lation have now accumulated. However, the only re-
cent review of the escalation literature (Sleesman et al.,
2012) focused largely on psychological, individual-
level aspects of the phenomenon—with only a fraction
of the review highlighting its contextual features. Meta-
analyses require quantitative findings to calculate
effect sizes and also a sufficient number of studies that
examine each bivariate relationship (Schmidt &
Hunter, 2015). Because studies of social and structural
determinants of escalation have often used qualitative
data and are rather fragmented in their focus, many of
these studies were not included in this recent meta-
analysis. As a result, less than 10% of the studies in-
cluded in the meta-analysis focused on social and
structural determinants of escalation, even though
some evidence suggests these factors—which are
found in the decision’s contextual surroundings—are
even more important than microlevel influences
(Drummond, 1994). Consequently, we think it is time
to evaluate what the literature offers in terms of the
context of escalation to identify new research avenues
and push toward a more nuanced understanding of
this important behavior.
Importance of Context
Context refers to “stimuli and phenomena that
surround and thus exist in the environment external
to the individual, most often at a different level of
analysis”(Mowday & Sutton, 1993: 198). A contex-
tualized approach to research offers a number of
important benefits (Johns, 2006; Rousseau & Fried,
2001). It allows researchers to develop a deeper ap-
preciation of a phenomenon, allowing for new in-
sights and interpretations of when and how the
phenomenon develops. For example, does the pro-
cess of escalation unfold in different ways depending
on the attributes of leaders of a team or organization
and by their level of involvement in the decision
process? By focusing on context, researchers can also
discover relationships between variables that may
otherwise go unnoticed or misinterpreted. For in-
stance, how might the availability of resources and
performance of the overall organization influence
the likelihood of escalation? Finally, an appreciation
of context allows for integration across research
areas to weave various findings into a cohesive pic-
ture of a literature. To illustrate, what are the com-
monalities that emerge when considering escalation
across multiple levels of analysis?
Although organizational researchers have con-
ceptualized context in different ways, most have
emphasized multiple levels of analysis (e.g., Johns,
2006; Mowday & Sutton, 1993; Rousseau & Fried,
2001). A multilevel perspective allows researchers
to develop more robust theories to better capture
the complexity of organizations, increase their
predictive power, and boost the real-world rele-
vance of research (Hitt, Beamish, Jackson, &
Mathieu, 2007; Kozlowski & Klein, 2000). Further-
more, examining the context through a multilevel
lens has the potential to bridge the micro and macro
divide in management research (Bamberger, 2008).
As such, we employ a multilevel framework in our
review and analysis of research on the context of
escalation.
Our Approach
In terms of the scope of our review, we cast a wide
net by reviewing escalation research in management
and related disciplines. Specifically, during fall of
2016, we searched the ISI Web of Science database
for academic articles using the following keywords:
escalation,de-escalation,sunk cost,project com-
pletion,completion bias,incremental investment,
entrapment, and persistence. To keep our collection
of articles to a manageable size, we restricted our
search to the 50 journals listed in the 2016 edition of
the Financial Times list of top business journals,
a metric commonly used in business school rank-
ings. We acknowledge that some articles in the
literature might inevitably be overlooked, but we
aimed to strike a balance between the breadth and
depth of analysis. We examined each article for
relevance in terms of whether it examined the context
of escalation of commitment according to the multi-
level conceptualization of context that we discussed
earlier (e.g., Johns, 2006; Mowday & Sutton, 1993;
Rousseau & Fried, 2001). Specifically, we included
a given article in our review if it offered insights
about escalation behavior beyond the individual
2018 179Sleesman, Lennard, McNamara, and Conlon

level (i.e., if the article focused on group, organizational,
or external factors). As illustrated in Figure 1, we
structure our article by level of analysis, each orga-
nized by dominant themes that emerged. Tables 1–3
provide a list of the sources that contributed to
our review.
We begin by focusing on proximal influences on
escalation decisions, namely research related to
group context. We then move up to the organiza-
tional level of analysis to uncover the wide variety of
forces at play when escalation decisions are made in
organizations. Next, we examine influences that are
external to organizations, as they are the most distal
to escalation decisions. We employed this organiz-
ing structure because group and organizational
characteristics provide context for individuals’
behavior, and the external environment provides
context for the organization (Johns, 2006). In each
section, we identify a number of opportunities for
future research. Finally, we discuss some key com-
monalities that we identified across these levels as
well as a number of interesting new questions that
arise when examining escalation behavior across
multiple levels of context.
GROUP CONTEXT
Decisions are frequently made collaboratively
(Edmondson, 2012; Tannenbaum, Mathieu, Salas,
& Cohen, 2012)—and for a good reasons. Group
decision-making can offer many benefits, such as
error reduction and knowledge aggregation (Kerr &
Tindale, 2004; Larrick, 2016; Sunstein & Hastie,
2015). Specifically, when decisions integrate multi-
ple perspectives, the errors associated with the idio-
syncratic viewpoints of any one person can be offset
and corrected by others. In addition, groups offer
expansive knowledge across individuals, which al-
lows them to generate a fuller understanding of
a given situation. However, groups do not always
yield better decisions than individuals, and there has
been extensive debate and inconsistent findings
FIGURE 1
Overview of Each Level of Analysis and Dominant Themes Within Each
STAKEHOLDER
ENVIRONMENT
• COMMUNITY
• MEDIA
• STAKEHOLDER PROXIMITY
TO PROJECT
FORCES
•
MARKET FACTORS
•
INDUSTRY FACTORS
CULTURAL INFLUENCES
• CROSS-CULTURAL
DIFFERENCES
DECISION
ATTRIBUTES
• DECISION PROCESS
STRUCTURE
• FEEDBACK
INFORMATION
• DECISION TYPE
EXECUTIVE LEADERSHIP
ATTRIBUTES
• OVERCONFIDENCE
• MOTIVATION
• IDENTITY &
COMMITMENT
ORGANIZATION ATTRIBUTES
• ORG. IDENTITY & CULTURE
• ORG. STRUCTURE, AGE, &
PERFORMANCE
• POLITICAL ENVIRONMENT
• CORPORATE GOVERNANCE
COGNITIVE INFLUENCES
• OPTIMISM TRAPS
• PERSPECTIVE DIFFERENCES
RELATIONAL INFLUENCES
• RESPONSIBILITY EFFECTS
• AUTHORITY PRESSURES
External
Context
ESCALATION DECISIONS
Organizational
Context
Group
Context
MARKET & INDUSTRY
180 JanuaryAcademy of Management Annals

regarding whether groups amplify or attenuate
individual-level biases (Hinsz, Tindale, & Nagao,
2008; Kerr & Tindale, 2004).
Although research has discovered a number of
explanations for when and why groups vary in their
decision-making effectiveness, they generally re-
volve around two primary sets of influences: cogni-
tive and relational (Larrick, 2016). Cognitive
influences relate to how individuals share and
integrate informational resources with others to
generate collective decisions (e.g., Hinsz, Tindale, &
Vollrath, 1997; Nijstad & De Dreu, 2012), whereas
relational influences are concerned with interper-
sonal dynamics among individuals (e.g., Edmondson &
Lei, 2014; Janis, 1972). In keeping with this broad
conceptualization, we structure our review of group-
relatedresearchonescalationaccordingtothesetwo
categories.
Cognitive Influences
We identified two themes in our review of cogni-
tive influences: optimism traps and perspective
differences.
Optimism traps. Although individuals tend to be
overly optimistic, research suggests that groups are
even more so. For instance, when groups estimate
how long they will take to complete a task, discus-
sion promotes narrowed attention on information
about successful task completion—which reinforces
their optimistic outlook (Buehler, Messervey, &
Griffin, 2005). This is likely to be compounded in
complex and dynamic situations, which people
tend to reduce into simpler representations of
reality—often leading them astray (Cronin, Gonzalez, &
Sterman, 2009; March, 2006). In this vein, Van
Oorschot, Akkermans, Sengupta, and Van Wassenhove
(2013) analyzed how a new product development
team failed to realize a project was in trouble until it
was too late.They found that the team’s perceptions of
project performance were consistently biased be-
cause members used various “information filters”
that skewed their understanding of reality. In par-
ticular, they (1) allowed positive feedback to cloud
their comprehension of negative feedback, (2) oper-
ationalized underlying problems in different ways,
and (3) overestimated the workload that staff
members could perform. As a result, the team
unduly extended early stages of the project,
which decreased slack and time for work at later
stages, ultimately resulting in failure. Drawing
inferences from their qualitative inquiry, they
argued that because “small problems can snow-
ball into big problems”(Van Oorschot et al., 2013:
303), group escalation can be curtailed if mem-
bers work together to put their wishful thinking
aside and be attentive to even minor signs of
failure. The authors postulated that groups can
make better project decisions if members are
given feedback information about the decision
environment, others’perceptions, and the likely
effects of possible future decisions (Gonzalez,
2005).
Using a psychoanalytic lens, Fotaki and Hyde
(2014) described how collectives can develop “blind
spots”that prevent them from recognizing problems
and eventually commit them to failing strategies. In
their analysis of three cases from the National Health
Service in England, the authors found that when
projects are idealized (e.g., their success is tied to the
organization’s image or there is excessive pressure
from senior management to succeed), individuals
feel compelled to psychologically “split”any evi-
dence of failure from themselves and project blame
onto other groups or projects. In this manner,
“blaming distorts reality and is a substitute for
meaningful action to resolve problems”(Fotaki &
Hyde, 2014: 456). This self-protective process of
idealization, splitting, and blaming becomes socially
reinforced over time, allowing groups to maintain
a collective fantasy that projects are going well when,
in fact, they are not.
Perspective differences. Escalation decisions
made in groups also appear to be influenced by the
extent to which members diverge in their perspec-
tives of the given situation. Westphal and Bednar
(2005) considered how the group dynamics of boards
of directors could contribute to strategic persistence
TABLE 1
Articles on Group Context
Article
Cognitive
Influences
Relational
Influences
Bazerman, Giuliano, and Appelman
(1984)
X
Fotaki and Hyde (2014) X X
Keil and M¨
ahring (2010) X
Moon, Conlon, Humphrey, Quigley,
Devers, and Nowakowski (2003)
XX
Roberto (2002) X
Van Oorschot, Akkermans,
Sengupta, and Wassenhove (2013)
X
Weick and Sutcliffe (2003) X
Westphal and Bednar (2005) X
Whyte (1991) X
Whyte (1993) X
2018 181Sleesman, Lennard, McNamara, and Conlon

in response to poor firm performance. Using survey
data from over 450 directors from more than 250
corporations, they found that when faced with low
performance, board members tend to underestimate
how much others share their concerns, leading them
to avoid expressing their opinions during board
meetings. As a result of this perspective-taking fail-
ure (“pluralistic ignorance,”cf. Miller & McFarland,
1987), boards can fail to initiate strategic change
when it is needed most. Interestingly, these negative
effects were reduced if the friendship ties among
directors were dense and if boards were de-
mographically homogenous, presumably because
members were more aligned (perceived or other-
wise) in their concerns about firm performance, or
because perspective-taking was more accurate.
A study by Moon, Conlon, Humphrey, Quigley,
Devers, and Nowakowski (2003) provides additional
evidence that the different perspectives held by
group members can influence escalation behavior.
They compared groups that made escalation de-
cisions as a whole with groups whose members pri-
vately considered the escalation decision before
making the decision again as part of a group. The
authors argued that the divergent escalation per-
spectives formed by individuals in the latter condi-
tion would lead to compromises during group
discussions, resulting in continuing the failing project
at some level of funding (“incremental”investments),
rather than taking drastic measures such as project
abandonment. Their argument was supported by re-
sults from a laboratory study which also found that
prior-consideration groups reported being more
likely to escalate but also less confident in their
decision-making than the no-prior-consideration
groups, suggesting that group escalation decisions
TABLE 2
Articles on Organizational Context
Article Decision Attributes Executive Leadership Attributes Organization Attributes
Booth and Schulz (2004) X
Boulding, Morgan, and Staelin (1997) X
Branzei, Bryant, Vertinsky, and Zhang (2004) X
Buchholtz, Lubatkin, and O’Neill (1999) X X
Chng, Rodger, Shih, and Song (2012) X X
Colombo and Delmastro (2001) X
DeTienne, Shepherd, and De Castro (2008) X X
Drummond (1995) X
Fotaki and Hyde (2014) X
Furneaux and Wade (2011) X
Gimeno, Folta, Cooper, and Woo (1997) X
Guler (2007) X
Hayward and Shimizu (2006) X
Hoang and Gimeno (2010) X
Inkpen and Ross (2001) X X X
Jacobides (2007) X
Joseph and Gaba (2015) X
Keil (1995) X
Keil and M¨
ahring (2010) X X
Keil and Montealegre (2000) X
Lowe and Ziedonis (2006) X
Maslach (2016) X
McCarthy, Schoorman, and Cooper (1993) X
McNamara, Moon, and Bromiley (2002) X X
Miller and Sardais (2013) X
Noda and Bower (1996) X
Park, Westphal, and Stern (2011) X
Roberto (2002) X
Ross and Staw (1986) X
Ross and Staw (1993) X X
Sinha, Inkson, and Barker (2012) X
Staw, Barsade, and Koput (1997) X
Weick and Sutcliffe (2003) X X
Westphal and Bednar (2005) X
182 JanuaryAcademy of Management Annals

were affected by the varying opinions that surfaced
during discussions.
Perspective differences may stem not only from
inferences about other members or the private
consideration of decisions before group discus-
sions, but also from the introduction of new mem-
bers to a group. In a pair of case studies (a deposit
system being developed by Eurobank and a data-
base systems integration in California), Keil and
M¨
ahring (2010) noted how opinion differences
among key project team members resulted in
“project drift”and the continued commitment of
time and financial resources, despite growing evi-
dence of project failure. Eventually, members of the
information technology (IT) and business develop-
ment units started to notice multiple pieces of bad
news about the project (problem emergence); this
information, highlighting the need for substantial
additional expenditures and time for project
completion, was then communicated to steering
committees overseeing the project (problem visibility).
Important in this last stage was the role played by
external parties who were often recently added to the
project teams. The fresh perspectives from these
new information sources often provided negative
reviews that, along with their perceived objectivity,
facilitated de-escalation in the form of significant
project alteration (in the case of Eurobank) or proj-
ect termination (in the case of the California data-
base integration).
Potential for future research. Although these
studies have identified a number of interesting
findings regarding the cognitive influences on esca-
lation in groups, much more work is needed. First,
we urge researchers to develop a better under-
standing of how group diversity influences escala-
tion behavior. A large body of work has found that
diversity enables groups to leverage informational
resources and benefit in many ways, including en-
hanced decision-making, problem-solving, and cre-
ativity (van Knippenberg & Mell, 2016). Bringing
multiple perspectives to the table may help groups to
develop a more comprehensive understanding of an
escalation situation before taking action. For in-
stance, Dilts and Pence (2006) examined how eval-
uations of project failure can differ depending on
one’s role on the project. Specifically, they found
that project managers emphasized operational
factors (e.g., changes in project complexity and
completion time), whereas executives were more
interested in the “big picture”(e.g., whether the
customer’s needs are satisfied). Thus, structural
differences among individuals can result in a richer
body of information that may feed into escalation
decisions.
However, the Westphal and Bednar (2005) study
discussed earlier suggests that diversity actually
contributes to escalation of commitment. Recall
their finding that members of demographically
heterogeneous boards of directors underestimated
the extent to which others shared their performance
concerns and were less likely to voice opinions,
ultimately leading to the avoidance of strategic
change. Of course, it is premature to conclude that
diversity accelerates the tendency of groups to es-
calate, but more work is clearly needed in this area,
especially given that teams in the workplace are
becoming increasingly diverse (Jackson & Joshi,
2011).
In addition, because escalation of commitment can
be viewed as a bias rooted in information processing
(Schultze, Pfeiffer, & Schulz-Hardt, 2012), we rec-
ommend that scholars consider group-level informa-
tion processing (De Dreu, Nijstad, & van Knippenberg,
TABLE 3
Articles on External Context
Article
Stakeholder
Environment
Market &
Industry
Forces
Cultural
Influences
Davidsson and
Gordon (2015)
X
DeTienne, Shepherd,
and De Castro (2008)
X
Devigne, Manigart,
and Wright (2016)
X
Greer and Stephens
(2001)
X
Hsieh, Tsai, and Chen
(2015)
X
Keil (1995) X
Keil and Montealegre
(2000)
X
Keil, Tan, Wei,
Saarinen,
Tuunainen, and
Wassenaar (2000)
XX
Lant, Milliken, and
Batra (1992)
X
Montealegre and Keil
(2000)
XX
Ross and Staw (1986) X
Ross and Staw (1993) X
Sharp and Salter
(1997)
X
Sinha, Inkson, and
Barker (2012)
X
Walker (2000) X
Zardkoohi (2004) X
2018 183Sleesman, Lennard, McNamara, and Conlon

2008; Hinsz et al., 1997) as a broad theoretical
framework for understanding the cognitive processes
involved in escalation in groups. Consider, for ex-
ample, how epistemic motivation might explain dif-
ferent relationships found in our review. Groups
whose members are epistemically motivated (repre-
sented by personality dispositions such as need for
cognition, or by situational features such as process
accountability; see De Dreu et al., 2008 for several
others) tend to process information more deeply,
leading to improved decision-making and a more ac-
curate understanding of a situation. The introduction
of new members into a group may raise the group’s
level of epistemic motivation and thereby increase the
probability of de-escalation (Keil & M ¨
ahring, 2010)
because it leads members to think more critically
about the decision at hand. By contrast, “information
filters”faced by project teams (Van Oorschot et al.
2013) may suppress epistemic motivation levels and
produce a superficial processing of information
whereby positive information “crowds out”negative
feedback, leading to escalation. As these examples
illustrate, group information processing may be
a useful lens from which to understand escalation
in the context of groups, and we encourage the
integration of these literatures.
Relational Influences
Next, we illuminate work that has investigated
relational aspects of escalation in the context of
groups. Two dominant themes emerged in our re-
view of this domain: responsibility effects and
authority pressures.
Responsibility effects. A classic explanation of
escalation is that under the weight of responsibility
forhavinginitiatedacourseofaction,individuals
later feel compelled to justify their decision by
escalating commitment to it (Brockner, 1992; Staw,
1976). In some cases, groups act the same way.
For example, Bazerman, Giuliano, and Appelman
(1984) suggested that groups would collectively
experience pressures to justify past decisions. Using
an investment-based scenario task, they manipu-
lated social unit (individuals versus four-person
groups) and decision responsibility (high: the ini-
tial course of action was explicitly chosen or low: the
course of action was assigned to them). Results in-
dicated that high-responsibility individuals and
groups escalated to a greater extent than low-
responsibility individuals and groups, but they
found no significant differences in escalation be-
tween individuals and groups.
In contrast, Whyte (1991) argued that in a group
decision-making context, self-justification concerns
are diminished because personal responsibility for
the initial decision becomes diffused across group
members, resulting in attenuated levels of escalation
(cf. Leatherwood & Conlon, 1987). In a laboratory
study, he randomly assigned participants to one of
the following three conditions that manipulated
locus of responsibility, from high to low: individual
(participants were described as having personally
made the initial decision), group (participants were
described as being part of a group that made the
initial decision), or control (participants were told
that another individual made the initial decision).
As expected, these varying degrees of responsibility
were associated with different levels of escalation:
Participants in the individual condition were more
likely to escalate, and escalate to a greater degree,
than those in the group condition, followed by those
in the control condition.
Although the previously mentioned studies fo-
cused on decision responsibility spread across group
members, others examined how groups react when
such responsibility was imbued in group members
themselves. Whyte (1993) conducted a subsequent
study that was similar to the Bazerman et al. (1984)
design, except that participants in the group condi-
tion considered the escalation decision on their own
before group discussion (similar to the Moon et al.
[2003] study discussed earlier). Although a seem-
ingly minor change, prior consideration by in-
dividuals was argued to result in pronounced levels
of group escalation because it would engender a stron-
ger sense of personal responsibility and motivation
for members to defend their perspectives to others
during group discussions (cf. Brockner, Rubin, &
Lang, 1981). Both the Bazerman et al. (1984) and
Whyte (1993) studies found that such groups
were more likely to escalate than individuals mak-
ing decisions independently. As we noted earlier,
Moon et al. (2003) also found that such groups
demonstrated more escalation behavior than
groups whose members did not engage in prior
consideration.
Combined, this work suggests that groups are more
likely to escalate when group members take on
personal decision responsibility and less likely to
escalate whenpersonal responsibility is spread across
multiple members. Interestingly, the relationship
appears to change when groups do not see them-
selves as working collectively as one decision team,
and instead see group members as representing
disparate organizational units. Recall the earlier
184 JanuaryAcademy of Management Annals

discussed study of projects in the English National
Health Service (Fotaki & Hyde, 2014), which
revealed that decision-makers working in a de-
cision team that crossed organizational boundaries
appeared to perceive less personal decision re-
sponsibility and assigned responsibility for de-
cision failures to other parts of the organization.
Importantly, this self-protection form of diffusion of
responsibility resulted in the creation of collective
“blind spots”—which resulted in the continuation
of failing strategies rather than a search for un-
derlying problems. In other words, it seems that
groupsendedupescalatingcommitmentbecause
group members eschewed personal decision re-
sponsibility. Clearly, moreworkisneededtobetter
understand how responsibility effects influence
group escalation decisions.
Authority pressures. Another set of studies fo-
cused on how authority pressures can quash dissent
or influence group norms, resulting in escalation of
failing projects. Weick and Sutcliffe (2003) studied
how England’s Bristol Royal Infirmary became
entrapped into actions that fostered repeated cycles
of behavior that sustained low performance in its
pediatric cardiac surgery program. The authors
noted that key decision-makers at the hospital (no-
tably, surgeons and the CEO) had a significant
amount of autonomy, which allowed them to
construct and project a façade that dismissed poor
performance indicators and compelled others to do
the same. Despite swelling mortality rates, scores of
formal complaints, external investigations, and ar-
ticles criticizing their patient care, they rejected this
negative feedback and maintained commitment to
failing practices. In addition, these high-status in-
dividuals were described by peers as being dismis-
sive, autocratic, and intimidating. Such leadership
styles fed a culture of silence that stifled possibilities
to learn from failure and potentially abandon failing
practices.
Similarly, Roberto (2002) analyzed the events that
led to the death of five climbers during a Mount
Everest trek in 1996. After experiencing a number of
setbacks, the group realized that they were not going
to reach the summit before the predetermined turn-
around time that would have enabled them to de-
scend safely down the mountain. However, survivor
interviews and written accounts suggest that the
climbers did not want to relinquish their sunk costs
(time and financial expenditures), and as they got
closer to the summit of Everest, it became even more
difficult to stop. In his analysis of the tragedy,
Roberto (2002) identified authority pressures as
a principal source that made stopping especially
challenging. For instance, some climbers did not feel
comfortable voicing their concerns to group leaders
because of perceived status differences (in terms of
expertise). Finally, authoritarian leadership dis-
couraged group members from voicing their dis-
senting views. In a provocative illustration of this,
one of the group leaders declared before the trip that
he would not tolerate any discord from group mem-
bers, as his “word will be absolute law”(Roberto,
2002: 145). These factors culminated in the group
escalating commitment to achieving their (in-
creasingly unwise) goal.
Potential for future research. Together, this work
highlights the powerful ways in which relational
factors can influence escalation decisions in groups.
In many cases, escalation decisions appeared to be
affected by groupthink, the desire for groups to strive
for consensus by minimizing conflict and suppress-
ing dissent at the expense of realistically appraising
alternative courses of action (Janis, 1972; Sunstein &
Hastie, 2015). Because groups appear to prefer
escalating commitment, especially when members
have individually considered the decision before-
hand (Moon et al., 2003; Whyte, 1993), preferences
for de-escalation are likely relegated to dissenters
who are up against powerful group pressures. High-
lighting the challenge of this task, Whyte (1993)
found that whereas 50% of the groups in his sample
ultimately escalated if an initial minority favored
escalation, only 2% de-escalated when an initial
minority favored de-escalation. Thus, dissenters
pushing for de-escalation appear to be fighting an
uphill battle in terms of swaying their group’s
opinion.
More work is needed to understand and discover
ways of strengthening the influence of de-escalation
preferences in groups. One fruitful avenue of re-
search may involve psychological safety—the
shared belief that the group is safe for interper-
sonal risk-taking (Edmondson & Lei, 2014). When
psychological safety beliefs are established in a group,
members feel free to speak up and go against the
status quo, and this becomes especially important
when groups are faced with uncertainty and com-
plexity (Edmondson & Lei, 2014)—two key features
of escalation situations (Bowen, 1987; Staw & Ross,
1987). Research should examine the extent to which
psychological safety may allow de-escalation argu-
ments to surface and result in a more accurate ap-
praisal of the situation (combining the relational and
cognitive perspectives of escalation in groups).
However, it is important to keep in mind that
2018 185Sleesman, Lennard, McNamara, and Conlon

psychological safety does not always result in pos-
itive outcomes. Pearsall and Ellis (2011) found that
psychological safety can create an environment that
permits members to pursue self-interested, and even
unethical, behaviors. Thus, if escalating results in
outcomes that benefit group members (in terms of
their reputation, authority, etc.), it is possible that
psychological safety might actually fuel the flames
of escalation in groups.
In addition, a number of the studies on relational
aspects of escalation in groups emphasized the im-
portant role of group leaders. It appeared that when
individuals had serious reservations about escalat-
ing commitment, their concerns were frequently
stymied by high-status individuals who discouraged
dissent (e.g., Fotaki & Hyde, 2014; Roberto, 2002;
Van Oorschot et al., 2013; Weick & Sutcliffe, 2003).
Domineering leadership is one of the most toxic
contributors to ineffective group decision-making
(Larrick, 2016; McCauley, 1998). Powerful in-
dividuals tend to dominate group discussions, pre-
cluding others from speaking their mind—and
consequently, group decisions and performance
may suffer (Tost, Gino, & Larrick, 2013). Beyond our
elaboration of the role of authoritarian leadership,
studies should explore other aspects of leadership
that may influence the willingness of members to
voice their concerns about escalation. For example,
differences in the quality of relationships between
leaders and followers (Erdogan & Bauer, 2014) or the
extent to which leaders are adaptive to the attributes
of group members or the given situation (DeRue,
2011) might also be related to the tendency for groups
to escalate versus de-escalate.
ORGANIZATIONAL CONTEXT
Because many decisions are embedded in organi-
zational settings, a sizable body of research has
examined the role of organizational factors that
exacerbate or attenuate the tendency of decision-
makers to escalate. To summarize this literature, we
focus on three key sets of factors within organiza-
tions that influence escalation behavior: decision
attributes, executive leadership attributes, and or-
ganization attributes. We center on these factors
because they allow us to (1) follow our overall
framework by considering both proximal and distal
factors that reside at the organizational level and
(2) focus on factors that influence decision choices
and performance.
We first discuss decision attributes since man-
agement research has for several decades revealed
the strong effect of decision routines and processes
on how decisions unfold and their likelihood of
success (Simon, 1947; Sutcliffe & McNamara, 2001).
Next, we examine executive leadership attributes,
which can have a strong effect on the courses of ac-
tion a firm undertakes (Staw, Barsade, & Koput, 1997;
Wowak, Mannor, Arrfelt, & McNamara, 2016) as well
as the resulting performance of the firm (Quigley &
Hambrick, 2015). Finally, we consider organization
attributes, as decisions can be heavily influenced
by the broader organizational context within
which decision-makers reside (McNamara, Moon, &
Bromiley, 2002).
Decision Attributes
Research has focused on a number of decision
characteristics that vary widely within and among
organizations. In particular, we discuss three factors:
how the decision process is structured within the
organization, the feedback information that is avail-
able, and the type of decision undertaken.
Decision process structure. Two key aspects of
decision process structure have been tied to escala-
tion behavior. First, there is evidence that setting clear
decision goals or rules for when to de-escalate before
embarking on the course of action can facilitate
de-escalation. In an experimental study of executive
MBA students making new product development
decisions, Boulding, Morgan, and Staelin (1997)
found that they were more likely to withdraw from
a failing project when they precommitted to a rule
about the conditions that would trigger de-escalation.
The authors argued that such a decision rule reduces
the ability of managers to ignore or distort relevant
information. Similarly, in a study of bank lending
decisions, McNamara et al. (2002) found that setting
clear levels at which a loan would be seen as too risky
facilitated de-escalation. Relatedly, in their casestudy
of two major IT projects, Keil and M¨
ahring (2010)
concluded that setting clear decision goals up front,
scheduling review points, and identifying predefined
“alarm bell”points when thresholds are crossed in-
creased the ability of firms to de-escalate once projects
begin to fail. In line with the argument offered by
Boulding et al.(1997), Keil and M ¨
ahring (2010)argued
that these decision attributes force decision-makers
to assess the state of a project and consider relevant
information when making a go/no go decision.
Second, research has found that bifurcating who
authorizes initial investments from who undertakes
decision progress reviews lessens the likelihood
of escalation. Arguing that this would reduce
186 JanuaryAcademy of Management Annals

self-justification pressures, Boulding et al. (1997)
demonstrated that when individuals did not make the
initial approval decision, they were more willing to
de-escalate a failing project at the review point
compared with settings where they were responsible
for both the initial decision and the ongoing review
of the project. Jacobides (2007) found a similar re-
lationship in his case study of a near-war between
Greece and Turkey in 1996. When decision authority
was centralized such that the same individuals were
responsible for initial assessments and ongoing de-
cisions, they used local cognitive framing, became
stuck in routine processes, and suffered from an
inability to reframe the situation, leading to pro-
nounced levels of escalation.
McNamara et al. (2002) also found that changing
decision authority facilitated de-escalation, but their
findings also offer a cautionary tale for organiza-
tional leaders. In the bank they studied, decision
authority switched to other decision-makers when
the risk of the loan reached a particular threshold.
Although this change in authority triggered de-
escalation, individuals facing the possibility of hav-
ing their decision authority taken away tailored
information about the loan to avoid having the
change in authority triggered, at which point they
increased their escalation of commitment to the loan.
In essence, managers doubled down on their prior
decision by understating the risk of the loan and es-
calated commitment in an effort to justify their ear-
lier choices and avoid having to admit a decision
failure.
In addition, research has demonstrated that the
organizational role of the individuals involved in
the decision decoupling can also affect tendencies
to escalate commitment. For example, Staw et al.
(1997) examined how turnover among different cat-
egories of senior bank managers influenced a bank’s
willingness to set aside money or completely write
off bad loans (in effect, de-escalating their commit-
ment). Their results showed that turnover of operating
executives was a stronger predictor of de-escalation
than turnover of board members, and that the
strongest turnover–de-escalation relationship oc-
curred when turnover involved outside succession.
It is likely that these outsiders were less susceptible
to the generation of the blind spots or cognitive
biases noted in many of the studies we reviewed
earlier.
Feedback information. The feedback information
that is available in the organization is another im-
portant decision attribute. This feedback can come
from either internal reporting systems or external
market feedback, but regardless of the source, orga-
nizations typically set up systems and filters that
determine the feedback information that decision-
makers have access to and are instructed or
incentivized to consider. Research finds that this
feedback information can significantly influence
escalation behavior. First, feedback that is acquired
early in the course of action can be especially
impactful. This early information, if negative, ap-
pears to have the potential to lead decision-makers to
question a course of action before their commitment
becomes cemented. For instance, in a case study of
a legal practice, Drummond (1995) found that early
negative feedback had a significant influence on
a lawyer opening an office, causing her to re-assess
the viability of the practice. Similarly, Noda and
Bower (1996) conducted a case analysis of the tele-
phone service company US West, finding that early
negative feedback in the firm’s efforts in the cellular
telephone market significantly curbed the manage-
ment’s enthusiasm for the market. On the other hand,
there is evidence that early positive feedback re-
inforces the decision-maker’s commitment to the
course of action. Based on the pattern of action they
found with a second regional phone company (Bell
South), Noda and Bower (1996) concluded that early
positive feedback associated with their wireless
telephone business solidified their commitment to
the market, pushing the firm to expand rapidly in
spite of market uncertainties.
Second, the clarity of feedback that decision-
makers receive can also affect their likelihood of es-
calating or de-escalating. In addition to the early
timing of feedback, Drummond (1995) concluded
that clear and extremely negative feedback facili-
tated de-escalation because it offered undeniable
insights about the project’s prospects. In a case study
of the development of the Shoreham nuclear power
plant in New York, Ross and Staw (1993) concluded
that, in contrast to clear feedback, ambiguous feed-
back increased the degree to which decision-makers
escalated commitment to the failing project. Am-
biguous information offers decision-makers the op-
portunity to focus on self-enhancing actions that can
be used to justify the continuation of their preferred
course of action. Similarly, in their case study of two
alliances, Inkpen and Ross (2001) found that the
difficulty in accurately measuring alliance perfor-
mance led firm managers to persist even though the
alliances did not appear to be meeting performance
expectations.
Joseph and Gaba (2015) provide broader quanti-
tative evidence for the relationships found in the
2018 187Sleesman, Lennard, McNamara, and Conlon

case studies that examined the role of feedback. In
a study of cellular phone manufacturers’responses
to performance feedback, they found that clear and
consistent negative feedback lessened their likeli-
hood of persisting in the face of problems. By con-
trast, they found that ambiguous feedback, whereby
different performance feedback measures are not
highly correlated, increased their likelihood of per-
sistence. The authors argued that ambiguity offers
managers the opportunity to discount negative in-
formation because “ambiguous feedback may not
provide a definitive assessment of performance”
(Joseph & Gaba, 2015: 1963), which allows managers
to selectively focus on more favorable data to justify
continuance.
Decision type. The type of decision involved in
the escalation dilemma is also likely to be an im-
portant factor. Surprisingly, we were able to find
only one study that has investigated this possibility.
Maslach (2016) investigated the escalation implica-
tions of whether innovation efforts undertaken by
a firm are incremental (similar to their existing
technologies) or novel (unrelated to their existing
technologies). He argued that firms would be more
willing to interpret the feedback they received in
a way that emphasizes the learning potential with
incremental innovation efforts, thus stimulating
persistence to failing efforts. In line with this
argument, he used a sample of medical device
manufacturing firms and found that they tended
to persist in failing efforts if they were incremental,
rather than novel, in nature.
Potential for future research. Although these
studies offer important insights into the role of de-
cision attributes, there are several key issues to be
addressed. First, we have a limited understanding of
the decision process steps that are critical to avoiding
escalation. We have evidence that setting clear de-
cision goals and preidentified criteria can trigger
de-escalation, but the literature offers a limited
understanding of who should formulate these spec-
ifications and how to develop strong commitment to
them. We also know that periodic formal decision
reviews provide opportunities to evaluate a de-
cision, but it is unclear when these reviews should
occur, what information is needed, and how to keep
them from becoming pro forma exercises that have
little impact on decision courses. In addition, al-
though prior research has shown that bifurcating
initial decision authority from review authority
lessens self-justification pressures, no research has
examined whether or not organizations need to
change the review person from one review to the
next. For instance, it is an open question if giving
a positive review at one stage may create pressures
to escalate at later review points.
In addition, although clear, unambiguous data are
helpful in facilitating de-escalation, this creates
a challenge with strategic decisions because the data
regarding the progress of a product development ef-
fort, acquisition, or alliance are rarely unambiguous
(Haunschild, 1994). This is especially true early in
the project’s life, when the feedback information is so
powerful. Furthermore, unambiguous feedback at
this point may not be highly predictive of long-term
performance and it might lead to inadequate persis-
tence to the decision course. Future research should
strive to examine how the timing and structure of
feedback can be balanced to avoid both the likeli-
hood of escalation and inadequate persistence as
well as how the most appropriate timing and struc-
ture of feedback is contingent on decision attributes.
Executive Leadership Attributes
Our prior consideration of leadership focused on
its impact on group decision-making, but there is
also a body of literature on how executive leader-
ship influences whether firms will escalate com-
mitment. The overriding logic is that the leader sets
the vision and tone for the firm, which influences
the decision behavior by others in the firm.
Unsurprisingly, much of this research has focused
on entrepreneurial firms where leader influence is
likely to be most evident.
Overconfidence. The most widely examined as-
pect of executive leadership in our review is over-
confidence. Leader overconfidence has been argued
to increase the likelihood of escalation because it
reduces the degree to which individuals attend to
negative information, contributes to a sense that they
can overcome obstacles or setbacks, and enhances
their belief that they can achieve their goals (Chng,
Rodgers, Shih, & Song, 2012; Roberto, 2002). Fur-
thermore, when a leader is overconfident, this sense
of invincibility can permeate others around the
leader (Roberto, 2002). McCarthy, Schoorman, and
Cooper (1993) offered the earliest evidence of this
relationship using data from a longitudinal survey of
over 800 entrepreneurs. In line with their arguments,
they found that entrepreneurial overconfidence was
strongly related to the likelihood of escalation early
in a venture’s existence. However, they also found
that the influence of overconfidence actually de-
creased after multiple periods of negative perfor-
mance feedback.
188 JanuaryAcademy of Management Annals

Just as the overconfidence of top-level leaders can
permeate the decisions of others in the organization,
the beliefs and language of others can also influence
the leader’s overconfidence. For example, Park,
Westphal, and Stern (2011) studied how flattery
and opinion conformity behaviors performed by
managers and board members influenced CEOs.
Using survey data from CEOs, outside board mem-
bers, and a set of top managers with whom the CEOs
had communicated in the prior year, the authors
found that ingratiation acts by the latter two groups
toward the CEO were commonplace, and that such
ingratiation led CEOs to become overconfident.
Moreover, this overconfidence became an impedi-
ment to the CEOs’ability to adjust corporate strategy
in the face of declining firm performance. Thus, al-
though theoretical work on escalation (specifically,
a firm’s responsiveness to the need to divest) sug-
gests that more outside directors can reduce escala-
tion pressure (e.g., by divesting a failing asset, cf.
Buchholtz, Lubatkin, & O’Neill, 1999) this empirical
work suggests that if these outsiders are focused on
ingratiation toward the CEO, efforts to de-escalate
will be less successful.
Although not explicitly examining overconfidence,
two other studies involved related issues. A case
study by Miller and Sardais (2013) found that persis-
tence in a failing venture can occur when entrepre-
neurs adopt a positive mindset about the future. In
addition, Sinha, Inkson, and Barker (2012), in a case
study of a New Zealand airline, concluded that CEO
celebrity contributed to firm escalation. The authors
argued that celebrity CEOs are likely to be over-
confident because of prior successes and the acco-
lades they receive from analysts and the press, leading
to a desire to escalate commitment.
Motivation. In addition to overconfidence, re-
search has examined how the motivation of execu-
tive leaders can influence escalation. Gimeno, Folta,
Cooper, and Woo (1997) studied how intrinsic mo-
tivation influenced the willingness of entrepreneurs
to persist with their business in the face of low per-
formance. They argued that intrinsic motivation
enhances the degree of satisfaction an entrepreneur
draws from the work environment, which reduces
the focus on actual venture performance. In line with
their argument, they found that entrepreneurs who
were intrinsically motivated, measured by the de-
gree to which they received psychic income from
running their company, were more likely to have
lower performance thresholds at which they would
persist with the venture. Thus, although intrinsic
motivation is generally considered to be a positive
factor as it relates to work performance (Cerasoli,
Nicklin, & Ford, 2014), it may lead to a tendency to
overcommit to a failing course of action.
DeTienne, Shepherd, and De Castro (2008), in
contrast, focused on the extrinsic motivation of en-
trepreneurs and argued that this form of motivation
would lead them to be more dispassionate in their
assessment of venture performance and more willing
to exit if they received feedback suggesting they
would not achieve their personal goals. They found,
using survey data on a sample of 91 entrepreneurs,
that those higher in extrinsic motivation were less
influenced by factors that tend to trigger escala-
tion, including market dynamism, personal in-
vestment, and the lack of alternative options for the
entrepreneur.
Identity and commitment. Finally, a set of studies
focused on the identity and commitment of execu-
tive leaders. This work suggests that when leaders
identify with and become heavily involved in de-
cisions, they become socially bound to them and fear
the loss of face both within and outside the organi-
zation if they allow the organization to de-escalate.
In a case study of British Columbia’s hosting of
a World’s Fair, Ross and Staw (1986) found that
a prominent local politician was reluctant to with-
draw from hosting the fair because doing so, after
having championed it, would damage how others
perceived him as a leader. In a qualitative study of
multiple business investments, Inkpen and Ross
(2001) found a similar pattern and concluded that
top management initial involvement with and com-
mitment to a decision course impacted the degree to
which the firms persisted with failing alliances.
Branzei, Ursacki-Bryant, Vertinsky, and Zhang
(2004) found, in a survey of Chinese senior man-
agers, that firms whose leaders were committed to
environmental responsibility were more likely to
escalate commitment to green initiatives. They ar-
gued that the leader’s values influenced the catego-
rization of decision options in line with these values
as opportunities and led the firm to focus on positive
cues when evaluating these preferred options. Fi-
nally, in a theoretical analysis, Hoang and Gimeno
(2010) argued that the degree to which the identity of
entrepreneurs is tied to a venture will be positively
related to their persistence with the venture.
As a whole, these studies show that organizational
decisions have the potential to unduly reflect the
interests of senior managers and tend to fall prey to
their escalation tendencies when the leader’s per-
sonal identity or commitment becomes highly sa-
lient within the organization. At the same time, the
2018 189Sleesman, Lennard, McNamara, and Conlon

earlier discussed Weick and Sutcliffe (2003) study of
the Bristol Royal Infirmary found that uninvolved,
hands-off leadership hampered the hospital’s ability
to respond to unacceptable levels of service by
changing course. Thus, it seems that both over- and
underinvolved leadership can trigger escalation.
Potential for future research. There are a number
of unexplored issues with regard to the effects of
executive leadership on escalation of commitment.
First, the literature has examined a limited range of
leader attributes. Although there is evidence that
confidence influences the proclivity of decision-
makers to escalate, future research should draw
more fully from research on executive attributes
that have been shown to influence organizational
actions. For example, research has demonstrated
that the narcissism of leaders influences the degree
to which firms aggressively pursue high-risk cour-
ses (Chatterjee & Hambrick, 2011). Similarly, re-
search has found that the regulatory focus (Gamache,
McNamara, Mannor, & Johnson, 2015) and temporal
orientation (Nadkarni & Chen, 2014) of executives
influence firm risk-taking. Future research could
extend work on these constructs and offer insight
to the escalation literature by examining the degree
to which these leader attributes influence commit-
ment to failing endeavors. In addition, future stud-
ies could draw on research that has examined the
influence of broad personality traits on the ten-
dency of individuals to escalate. For example,
there is evidence at the individual level that the
tendency to escalate is predicted by traits such as
neuroticism (Wong, Yik, & Kwong, 2006) and as-
pects of conscientiousness (Moon, 2001), but it
is unclear if these relationships translate to the
firm level.
Second, research should examine how aspects of
executive leadership interact with other features of
the organization or its strategy. Evidence shows that
the effects of executive attributes on escalation be-
havior may be contingent on factors such as the
structure of compensation they receive or the firm’s
performance level (Chng et al., 2012). Relatedly,
much of the research examining the influence of
leader attributes has been in entrepreneurial firms.
It is unclear whether these relationships will be as
strong in more mature firms. Future research
should examine the degree to which factors, such as
organizational structure and size, moderate the
influence of leader attributes on escalation. For
example, the extent to which an executive’s over-
confidence contributes to escalation is likely at-
tenuated in large organizations where decision
authority may be held by individuals several levels
down from the CEO.
Organization Attributes
A sizable body of literature has examined a num-
ber of organization attributes that influence escala-
tion behavior. These include the organization’s identity
and culture; structure, age, and performance; po-
litical environment; and corporate governance
structure.
Organizational identity and culture. Escalation
is related to the extent to which the decision is tied to
the organization’s identity and culture. As these ties
become stronger, it becomes hard for firms to change
course for multiple reasons. First, de-escalating
presents challenges inside the firm if the action is
perceived to be inconsistent with the firm’s identity
or values, raising a sense of cognitive dissonance
(Hinojosa, Gardner, Walker, Cogliser, & Gullifor,
2016) as internal stakeholders struggle with the log-
ical inconsistency. In line with this argument,
Inkpen and Ross (2001) found that firm identity
made it very difficult for McDonnell Douglas to
cancel its alliance with the Chinese Ministry of
Aeronautics. Managers considered the alliance cen-
tral to the firm’s ability to sustain its identity as
a commercial airline manufacturer. As a result, they
surmised that ending the alliance would signal
abandonment of its mission, which led the firm to
persist with the alliance long after decision-makers
realized it would not pay off. Similarly, the Ross and
Staw (1993) nuclear power plant study concluded
that LILCO, the utility that built the plant, considered
it central to the corporate vision of the firm—severely
hampering its ability to abandon the plant. Keil
(1995) also found evidence consistent with this ar-
gument. In a case study of a large project within
a computer manufacturer, he found that an organi-
zational culture that emphasized heroic stories of
leaders who saved failing projects and valued norms
of consistency inhibited the willingness of managers
to change course.
Second, firms with a strong identity and culture
can overlook important warning signs if factors in-
volving change are discounted as irrelevant or out-
side the organization’s information scope. Weick
and Sutcliffe (2003) found that a culture emphasiz-
ing optimism, learning curves, and improvement
over time can lead individuals to downplay or ignore
information indicating a need for dramatic changes.
Similarly, Keil and Montealegre (2000) noted that
a culture that emphasized the passionate pursuit of
190 JanuaryAcademy of Management Annals

getting projects operational led the London Stock
Exchange to ignore warning signs that the develop-
ment of its Taurus trading system was doomed to
failure.
In contrast to the previously mentioned work,
Booth and Schulz (2004) took a different approach,
arguing that a firm’s culture can also lead to de-
escalation. Examining the ethical environment
within the firm, they argued that research focusing
solely on economic drivers overlooks how managers’
framing of project continuance decisions is influ-
enced by their ethical sensibilities. They further ar-
gued that a strong ethical environment lessens the
likelihood that managers will act opportunistically
in their self-interests. Instead, managers in this set-
ting were expected to act in the interests of their firm.
Using experimental data on decisions by 131 middle
managers, they found that individuals were more
likely to de-escalate in a strong, rather than weak,
ethical setting. Similarly, Keil and M¨
ahring (2010)
argued that firms can develop a culture that en-
courages problem disclosure to raise faster aware-
ness of problems and trigger discussions of whether
the firm should persist, and they found supporting
evidence using case studies. Thus, there is evidence
that organizational culture can both exacerbate and
attenuate escalation tendencies.
Organizational structure, age, and performance.
A set of studies has examined the influence of basic
organizational attributes on escalation of commit-
ment. In a study of Italian metalworking firms,
Colombo and Delmastro (2001) examined the influ-
ence of organizational structure on the willingness to
de-escalate. They found that when a given produc-
tion plant was part of a multiplant system, the indi-
vidual plant faced a greater likelihood of
closure—suggesting that it is easier to de-escalate
when a given plant (or investment, more broadly) is
treated overtly as one of a portfolio of decisions or
operations. In addition, Lowe and Ziedonis (2006)
revealed differences in escalation tendencies between
established firms and new (entrepreneurial) firms
using data from efforts to commercialize inventions.
They found that established firms were more likely to
cancel licensing agreements than entrepreneurs, con-
cluding that the former were more likely to de-escalate
because they perceived the licenses to be part of a broader
investment portfolio they were undertaking—giving
them greater latitude to cancel the agreements if they
were not meeting expectations.
Relatedly, research suggests that when decisions
are embedded as part of an organization’s overall set
of investments and routines, escalation is more likely
to occur. In a survey-based study of IT executives,
Furneaux and Wade (2011) found that firms often
had difficulty extracting themselves from IT systems
that outlived their useful life spans when they were
connected to a range of other organizational in-
vestments and processes. Combined, these studies
demonstrate that the willingness or ability of firms to
de-escalate is influenced by the extent to which the
firm’s portfolio of decisions are interdependent,
resulting in each decision being embedded in a net-
work of other organizational activities. As the in-
terdependence of decisions increases, the discretion
held by any single decision-maker or group decreases,
as withdrawing would have to take into account the
broader fallout throughout the organization and
would potentially require cross-unit negotiations. In
essence, it appears that whether firms have a portfolio
of decisions influences their likelihood of de-
escalation, but the direction of the effect is contin-
gent on whether the failing action is interdependent
with the firm’s other activities.
Other studies have also found that strong prior
organizational performance can exacerbate escala-
tion tendencies. The logic is that prior success leads
decision-makers to develop a stronger sense that
they can overcome challenges with current initia-
tives (DeTienne et al., 2008; Keil, 1995). DeTienne
et al. (2008) argued, “[p]reviously successful expe-
riences may cause entrepreneurs to believe that they
have the ‘formula’correct and that persistence will
lead to success”(p. 535). Support for this argument
has been found in a case study of a computer man-
ufacturer (Keil, 1995), in technology firms (DeTienne
et al., 2008), and with commercial bank lending
(McNamara et al., 2002). These findings indicate that
a history of success leaves decision-makers more
resistant to data suggesting that the organization
needs to change course.
Hayward and Shimizu (2006) offered a contrasting
argument and set of findings. In a study of large
United States firms that had acquired other firms,
they demonstrated that strong prior performance
would make firms more willing to divest poorly
performing units. Their reasoning was that execu-
tives of well-performing firms could couch the fail-
ure of one unit in light of the firm’s overall strong
performance, leaving them more willing to take the
(publicly visible) action of de-escalating by divesting
the poorly performing unit. Thus, strong overall
performance offered them the cover needed to make
painful admissions of failures that could be com-
partmentalized. In contrast, executives of poorly
performing firms would want to avoid the attention
2018 191Sleesman, Lennard, McNamara, and Conlon

a divestiture would bring and the resulting height-
ened scrutiny they would receive.
Political environment. The political environment
of an organization can also influence whether
decision-makers persist when problems arise with
projects. Keil (1995) argued that political fights be-
tween organizational units can exacerbate escalation
tendencies because competing units will typically
attribute blame to others and persist to justify their
beliefs and force the organization to fix the other unit
to correct the problem. He found supporting evi-
dence from his analysis of a large computer manu-
facturer, where a strong rivalry between sales and
manufacturing employees led the firm to persist long
after it seemed inevitable that the software the firm
was developing would fail.
Similarly, Fotaki and Hyde (2014) found that the
degree to which decision-makers could assign blame
to other units leads to both greater persistence and the
inability to have meaningful, collective efforts to im-
prove a failing situation. In a study of venture capital
firms, Guler(2007) found that a political environment
where venture capitalist (VC) professionals had to
collectively vote to fund investments led to deal-
making and trading of support for initiatives each VC
had advocated. As a result, once a commitment was
made, it was difficult for any VC professional to leave
an investment in which he or she was a supporting
(rather than the championing) investor because doing
so could trigger retaliatory behavior whereby others
could withdraw from supporting the initiatives the
focal VC professional had championed.
Corporate governance. The final organizational
attribute found to relate to escalation is the corporate
governance of the firm. In a theoretical paper exam-
ining board structure, Buchholtz et al. (1999) argued
that firms whose boards had a higher proportion of
qualified outside directors and higher levels of stock
ownership by outside directors would be more likely
to de-escalate—highlighting the potential benefits of
active board members bringing differing perspec-
tives and experiences. However, these purported
benefits may not always bear out. Recall the
Westphal and Bednar (2005) study of corporate
boards, which found that the extent to which boards
were demographically heterogeneous led to greater
strategic persistence in the face of low firm perfor-
mance. These studies demonstrate the powerful and
complex influence that boards can have on the es-
calation behavior of firms.
Prior work has also considered the influence of top
manager compensation. In their theoretical analysis,
Buchholtz et al. (1999) argued that incentive
compensation would enhance the degree to which
executives are driven by extrinsic motivation, lead-
ing them to be more willing to de-escalate if they
believe it will enhance the value of their stock option
portfolios. Chng et al. (2012) found support for an
effect from incentive compensation but also dem-
onstrated that it was contingent on managers’core
self-evaluation. Using a sample of 216 Chinese MBA
students engaging in a decision simulation, they
found that for individuals high in core self-evaluation,
the presence of incentive compensation made them
persist more when they were not succeeding.
Potential for future research. There are at least
three potential research paths to help tease out un-
certain or mixed effects in this area. First, prior re-
search has found inconsistent effects for firm
performance. On the one hand, we see evidence that
strong performance leaves managers more confi-
dent in their abilities, which increases the likeli-
hood of escalation (DeTienne et al., 2008; Keil,
1995). On the other hand, research suggests that
strong performance increases the willingness of
decision-makers to de-escalate (Hayward and
Shimizu, 2006). Future research should examine
factors that may explain the differing consequences
of performance. For example, these mixed findings
suggest that the degree of external scrutiny man-
agers face may affect how performance influences
escalation, with higher scrutiny potentially in-
creasing the likelihood that strong performance will
trigger de-escalation.
Second, the effect of incentive compensation on
escalation tendencies remains unclear. Whereas
Buchholtz et al. (1999) theorized that incentive com-
pensation would lead to de-escalation because it
would enhance decision-makers’extrinsic motivation,
Chng et al. (2012) found that for managers high in core
self-evaluation, incentive compensation led to greater
escalation. However, it is worth noting that their
incentives were bonus-based rather than stock-based.
As a result, there was no downside risk when they were
not succeeding; they simply would not receive their
bonuses, incentivizing them to persist to achieve their
targets and obtain the bonus. With stock-based
incentives, especially restricted stock and stock own-
ership, executives face continued downside risk
(Devers, McNamara, Wiseman, & Arrfelt, 2008) be-
cause persisting could lead to further erosion of their
portfolio value, possibly producing a different re-
lationship between incentive compensation and
persistence. Thus, research should examine the influ-
ence of different forms of incentive compensation on
escalation.
192 JanuaryAcademy of Management Annals

Finally, it would be beneficial for research to include
multiple organizational attributes. For example, differ-
ences in executive leader attributes or the way decisions
are structured may moderate the extent to which de-
cision embeddedness influences escalation behavior.
Similarly, the time orientation of leaders may influence
the way in which feedback influences their escalation
preferences, with managers who are past-focused likely
being more affected by decision feedback relative to
those who are future-focused (Gamache & McNamara,
2017). In short, organizational attributes are clearly
powerful determinants of the extent to which decision-
makers fall prey to escalation, but research has only
begun to scratch the surface on these factors.
EXTERNAL CONTEXT
In addition to group and organizational factors,
escalation decisions are affected by sources outside
the organization. Insights can be found from a num-
ber of perspectives, including the stakeholder envi-
ronment, market and industry forces, as well as
cultural influences. As we did for the other sections,
we structure our analysis of external context by
moving from proximal sources (direct stakeholders
surrounding the organization) to more distal sources
(market, industry, and cultural factors).
Stakeholder Environment
Research has shown that not only decision-makers
inside a given organization but also those outside of
it can affect escalation. We will review work that
focuses on stakeholders from a number of perspec-
tives, including those in the community surrounding
the organization, the media, and the proximity of
stakeholders to the failing project.
Community. A collection of studies finds that es-
calation can result from projects becoming tied to the
community in which an organization resides. In their
analysis of a World’s Fair that we noted earlier, Ross
and Staw (1986) outlined an institutional explana-
tion for escalation. They concluded that as the proj-
ect unfolded, it became embedded in a larger
narrative and set of actions outside the organization.
For instance, the fair became emblematic of
Vancouver’s growing stature as a globally prominent
city on par with Toronto and as a city that completed
projects to which it committed. As planning pro-
gressed, the project also became tied to regional
infrastructure development and commitments by
private firms (e.g., hotels, restaurants, transportation
service providers, and small firms, such as souvenir
shops and photographers) to invest in and support
the fair.
Similarly, in a case study of a nuclear reprocessing
plant in the United Kingdom, Walker (2000) found
that the project became embedded in a set of legal
commitments with suppliers, contractors, utilities,
and politicians (promises to invigorate the regional
economy), hampering the government’s ability to
terminate the project once it began to struggle. In
addition, using logic derived from the economic
problem known as the “tragedy of the commons,”
Zardkoohi (2004) argued for the importance of ex-
ternal stakeholders in analyzing firm-level escala-
tion decisions. He suggested that firms are likely to
escalate their commitment to failing projects to the
extent that costs are passed on to parties outside the
firm, such as contractors, customers, or the public.
Thus, decision-makers appear to be more comfort-
able escalating commitment if the downside risk is
shifted to outside stakeholders.
Although external stakeholders can pose chal-
lenges for de-escalation, additional evidence by
Montealegre and Keil (2000) shows how external
stakeholders can also contribute to de-escalation.
This case study explores the automated baggage
system troubles at the new Denver International
Airport. The case plays out similarly to a multiparty
negotiation whereby different parties (e.g., the
mayor, representatives for the failed computerized
baggage system, vendors waiting to open at the air-
port, the Federal Aviation Administration, and an
external consulting company) were involved in
presenting information that led to sensemaking
about the cost overruns and delays, ultimately
(though perhaps not quickly enough) leading to an
exit strategy of abandoning the computerized system
in favor of a more traditional one. Montealegre and
Keil (2000) also suggested that decision-makers can
withdraw from failed projects more easily if they
appeal to the interests of outside stakeholders (see
also Ross & Staw, 1993).
Media. The escalation behavior of firms also ap-
pears to be influenced by narratives constructed by
the media. For example, Sinha et al. (2012) examined
the influence of the media as an important stake-
holder who helped Gary Toomey, the CEO of Air
New Zealand, hold onto his narrative (i.e., escalate
his commitment) as he sought to acquire another
airline, Ansett Australia. The case describes the
process of how the media first supported Toomey’s
acquisition behaviors—possibly reinforcing Toomey’s
decisions—but then later, partly in response to the
involvement of other parties (financial intermediaries
2018 193Sleesman, Lennard, McNamara, and Conlon

and other stakeholders), came to view the celebrity
CEO’s actions as misguided. Given that the media
can rapidly create and sustain narratives over
time (Bishop, Trevino, & Gioia, 2014), decision-
makers looking for external cues about escalation
can find it especially difficult to terminate their
projects.
Stakeholder proximity to project. In addition to
investigating the various stakeholders that influence
escalation decisions, we believe it is important to
take a step back and examine the role of the stake-
holders’proximity to a failing project. Research has
demonstrated that whether or not stakeholders re-
side close to the project can affect their reaction to
it. Devigne, Manigart, and Wright (2016) tracked
the investment decisions of 1060 venture capital
investments in 684 European companies, finding that
investors were considerably more likely to escalate
failing domestic investments than international
ones. The authors suggested this difference can “be
explained by cross-border investors having a lower
social and emotional involvement with the project
and a lower embeddedness in the local economic
and social environment, decreasing individual de-
cision biases”(p. 253). For instance, the geo-
graphically closer the venture capital investors were
to the project site, the more frequently they inter-
acted with the project entrepreneurs. This led them
to become embedded in the project, experience
normative pressures, and become emotionally
involved—making it harder to cut ties. The results
also suggested that venture capital investors were
less susceptible to concerns about damaging their
reputation when deciding to discontinue funding for
a cross-border investment than a domestic one.
Overall, the study indicates that distant projects
are evaluated more objectively than local ones. Thus,
contrary to the prevailing assumption in the in-
vestment literature that international venture capital
firms face a “liability of foreignness”(Zaheer, 1995),
the study suggests that international firms may ac-
tually have an advantage over domestic firms. As
many as one-third of venture capital investments
end up failing (Puri & Zarutskie, 2012), so this is
a particularly important area for escalation re-
searchers to explore.
Potential for future research. We think it is im-
portant that future research does more to examine
the role of external stakeholders, as this initial work
suggests they can have profound effects on escala-
tion. For example, research generally suggests that
escalation is likely to occur to the extent that the
project is embedded within a larger environment
(Ross & Staw, 1986; Walker, 2000). However, this
work also suggests that the contributions and buy-in
from external stakeholders can be vital for a project’s
success (Montealegre & Keil, 2000). Therefore, future
research should examine the seemingly paradoxical
relationship between the support (and guidance)
that external stakeholders provide and the intract-
ableness that their support can bring when the proj-
ect begins to decline. Future work should also
investigate the extent to which having stakeholder
support can cushion decision-makers from feeling
personally responsible for the potential failure of
a project.
Finally, research should more specifically exam-
ine the role of key financial stakeholders. For ex-
ample, activist investors who take a large stake in
a firm and push for immediate strategic changes may
be agents who can successfully trigger de-escalation
due to their strong influence (David, Hitt, & Gimeno,
2001). By contrast, escalation may be more prob-
lematic in firms that are primarily owned by dedi-
cated institutional investors (mutual funds and
pension funds that buy and hold firm stock for long
periods of time), as they are more likely to exhibit
patience when decisions do not offer a short-term
payoff (Connelly, Tihanyi, Certo, & Hitt, 2010).
Market and Industry Forces
Research has recently begun to examine the mar-
ket and industry forces that can affect an organiza-
tion’s likelihood to escalate commitment to failing
projects. Although this area is in need of further
development, particularly with regard to cross-
industry influences, we already see considerable
effects from a number of perspectives.
Market factors. DeTienne et al. (2008) examined
the role of market factors in entrepreneur decisions to
continue with underperforming firms. The authors
proposed three market factors—complexity, dyna-
mism, and munificence—that may affect escalation.
A complex market (e.g., with new and changing in-
formation) might be intimidating for a new entrepre-
neur to interpret and thus discourage them from
continuing with the venture, but an experienced
entrepreneur may be able to take advantage of the