UNIVERSITY OF
ILLINOIS LIBRARY
AT URBANACHAMPAIGN
BOOKSTACKS
F
CENTRAL CIRCULATION AND BOOKSTACKS
The person borrowing this material is re-
sponsible for its renewal or return before
the Latest Date stamped below. You may
be charged a minimum fee of $75.00 for
each non-returned or lost item.
Theft, mutilation, or defacement of library materials can be
causes for student disciplinary action. All materials owned by
the University of Illinois Library are the property of the Slate
of Illinois and are protected by Article 16B of Illinois Criminal
Law and Procedure.
TO RENEW, CALL (217) 333-8400.
University of Illinois Library at Urbana-Champaign
When renewing by phone, write new due date
below previous due date. L162
STX
BEBR
FA
PAPER NO. 1286
The Impact of Diffusibility of Blame on
Subsequent Persistence in a Project
Marya L. Leatherwood
Edward L. Conlon
College of Commerce and Business Administration
Bureau of Economic and Business Research
University of Illinois, Urbana-Champaign
BEBR
FACULTY WORKING PAPER NO. 1286
College of Commerce and Business Administration
University of Illinois at Urbana-Champaign
September 1986
The Impact of Dif fusibility of Blame on Subsequent
Persistence in a Project
Marya L. Leatherwood, Lecturer
Department of Business Administration
Edward J. Conlon
University of Iowa
Work on this paper was supported in part by NSF Grant
//BSN-8616275, "Resource allocation with sunk costs: A
behavioral approach." The authors thank Professor Barry Staw
his helpful comments on an earlier version of this paper.
Digitized by the Internet Archive
in 2011 with funding from
University of Illinois Urbana-Champaign
http://www.archive.org/details/impactofdiffusab1286leat
ABSTRACT
Although previous research on escalation and entrapment indicates that the
responses of resource allocators to setbacks may be related to
characteristics of the setback, that relationship has not been well
understood. Using a realistic resource allocation scenario, this study-
tested how one such characteristic, dif fusibility of blame, affected subject
responses. Withdrawal from a project occurred when blame could be diffused.
Conversely, continued involvement in the project resulted when blame for the
setback clearly rested on the allocator.
Over the last decade, a number of studies have demonstrated that when
courses of action result in setbacks or failures, the contextual
characteristics of both the decision to enter the course of action and the
decision to continue (or not continue) the course of action can affect the
amount of resources a decision maker will commit following the setback. In
the first study of this phenomenon, Staw (1976) showed that decision makers
who were responsible for the decision to enter a course of action tended to
escalate their commitment of resources following a setback. Responsibility
was studied because of its relationship to cognitive dissonance; that is, a
responsible decision maker would experience greater dissonance as a result
of the failure than one who was less responsible. Escalation was explained
as a way that decision makers would rationalize their failures and resolve
their dissonance.
Since Staw's first study, researchers have examined the effects of
additional context factors that interact with responsibility and failure to
affect decision makers' behavioral commitments to a course of action. These
factors are repetition of failure (Staw and Fox, 1977: McCain, 1986),
foreseeability of the failure (Staw and Ross, 1978), job insecurity (Fox and
Staw, 1979), involvement and visibility (Conlon and Wolf, 1980), prospective
information (Leatherwood and Conlon, 1986) , closeness to a goal (Rubin and
Brockner, 1975), group versus individual responsibility (Bazerman, Giuliano
and Appleman, 1984), presence of a model (Brockner, Nathanson, Friend,
Harbeck, Samuelson, Hauser, Bazerman and Rubin, 1984) and opportunity costs
and framing (Davis and Bobko , 1986; Northcraft and Neale , 1986) These
studies have primarily focussed on exogenous aspects of the social situation
(e.g., responsibility, visibility, groups) or endogenous characteristics of
the decision itself (e.g., opportunity costs and framing). Prompted by
results obtained by Staw and Ross (1978), this study examined how endogenous
characteristics of a setback, specifically diffusibility of blame to another
party, affected behavioral manifestations of commitment to a course of
action.
Effects of Setback Types
Staw and Ross (1978) found that allocations made following a setback
varied depending on the nature of the setback that had occured. In their
study, which utilized the "World Bank" case, subjects were given a choice of
three regions of Kenya in which to build an economic development project.
The information about each region contained a warning about a potential
problem. In the foreseeable setback condition, subjects were told that a
setback had occured and that it had been caused by the potential problem
mentioned in the information that had been provided about the region
selected. These problems were (1) corrupt local officials, (2) ineffective
economic incentives and (3) widespread illiteracy. In the non-foreseeable
setback condition the setback was attributed to excessive rainfall, a
possibility that subjects were not warned about.
Interestingly, Staw and Ross found that the amount allocated to the
project following the setback, a behavioral manifestation of commitment,
varied according to the type of setback. Generally, allocations were lower
following the endogenous setbacks (i.e., Responsible subjects withdrew from
the project), but there was -nsiderable variance in allocations among the
three setback conditions. Subjects allocated the least following the
corruption setback and the most following the illiteracy setback. The
amount allocated following the illiteracy setback was about the same as that
allocated following the (non-forseeable) rain setback.
In other studies of this type, allocators have been shown to escalate
rather than withdraw resource commitments in high responsibility contexts
(e.g. Staw, 1976; Staw and Fox, 1977; Fox and Staw, 1979). An important
difference between the Staw and Ross design and the others was the extent to
which the subjects were given explicit information about the cause of the
setback. Unlike the World Bank case, the other studies produced a setback
by indicating to the allocator that the performance of a previously funded
corporate division continued on a declining trend. No specific rationale,
such as corruption or rain, was provided for the decline. When no rationale
is given, it may be difficult for allocators to formulate specific
rationales on which to base their funding decisions. Research has shown
that they tend to continue funding following setbacks in such situations.
When rationales are provided, however, allocations may become part of the
allocator's means of rationalizing the setback.
The variation in allocations that occurred following Staw and Ross's
three foreseeable setbacks suggests that the ability to diffuse blame for
the setback to an outside party may be an important part of this
rationalizing process. For example, by withholding further funding to a
project beset by corruption, the allocator could claim that s/he was either
legitimately punishing the local officials or pressuring the local
government to purge those officials before further action could be taken.
In contrast, it is unclear that such a hard line would be considered
legitimate following the setb rk caused by illiteracy. The extent to which
behavioral commitment would increase as a result of setback foreseeability
would depend on the extent to which blame for the setback would be placed
entirely on the allocator. In the present study, we hypothesized that
setbacks which could legitimately be blamed on outside parties would lead to
different levels of commitment to the chosen course of action compared to
those which could not.
This study employed two variations on a single setback, a labor strike,
to investigate the diffusion hypothesis. In one case, the strike was a
legitimate response to an action taken by the allocator (i.e. the non-
diffusible condition), in the other it was not (i.e., the diffusible
condition). We hypothesized that when the setback was not foreseeable,-
therefore the responsibility of the allocator for the setback was low, the
placement of blame would not be a strong concern of the allocator and the
diffusibility of blame for the setback would not affect behavioral
manifestations of commitment. When foreseeability was high, behavioral
commitment would be greater following the setback for which blame could not
be diffused than following the one for which blame could be diffused. The
hypothesis, therefore, predicted an interaction between foreseeability and
diffusibility of the form outlined in Figure 1.
The Use of Multiple Dependent Measures
In previous studies, the extent of commitment to a project has been
inferred from the level of monetary allocation to that project. Even though
some of those studies produced results that were consistent with the
hypothesized commitment effects, allocations may be ambiguous indicators.
The problem is most easily seen in situations where the subject is provided
with the estimated development costs of the project as was the case in the
World Bank scenario (Staw and Ross, 1978; Conlon and Wolf, 1980) and in the
present study. When the amount necessary to complete a project is known or
can be estimated, allocations of that amount signify a clear decision to
continue, and allocations of zero indicate a clear desire to withdraw.
Allocations between zero and the cost estimate are difficult to interpret
since they could be attributed to an incremental approach to allocation
(e.g. Lindblom, 1959), an attempt to "control costs" or, more generally a
risk averse "wait and see" attitude. Similarly, allocations in excess of
the projection may indicate the subject's disbelief of the estimate or some
type of "adjustment" for potential cost increases.
The present study used three measures of behavioral commitment in erder
to clarify the allocators' motivations. The first measure (AMOUNT) was an
allocation of an amount which would be spent to further develop or complete
the project. Because the task used in this study was closely modeled after
a real context where cost projections are routinely provided, this measure
was expected to be distributed in a bimodal fashion with allocations
equalling either the remaining development costs or zero. As such, it was
expected to be a particularly good indicator of a desire to withdraw from
the project.
A second measure (OVERAGE) allowed individuals to set aside funds which
could be used in the event of further setbacks. It was explained that funds
in this account, if not used, would revert to the allocators' general fund
and could be used for future projects. The use of an overage amount
provided an indicator of commitment based on how willing decision makers
would be to exceed the projected budget in order to finish the project. It
was expected to be a good indicator of a subject's desire to continue, but
would not distinguish allocat ts who desired to withdraw from those who
desired to continue without committing additional resources. Finally, an
important alternative to the abandonment of projects which develop capital
assets (ie. land, buildings, financial instruments, etc.) is for managers to
seek to sell the assets and recover as much of the development cost as
possible (see Northcraft and Wolf, 1984). To measure this propensity,
allocators were asked to indicate their desire to sell the project (PRTYSEL)
on a four point scale. Unlike AMOUNT and OVERAGE, PRTYSEL can indicate
the desire to withdraw, continue or escalate resource commitment with about
equal sensitivity.
METHOD
Subjects
Sixty-seven business students volunteered to participate in what was
described as a decision making study. The members of this sample ranged in
age from 19 to 45 with a mean age of 26.7. Twenty- four subjects were MBA
students who were employed full time. These students had between 1 to 10
years experience in their current jobs (mean =2.9) and 1 to 19 years in their
current organization (mean - 6.2). The remaining subjects were in their
last 2 years of undergraduate study in business administration and had
completed the introductory courses in finance, accounting and economics.
Subjects were randomly assigned to experimental conditions.
Procedure
Subjects were provided a package of materials that contained the
following instructions:
This is an exercise in decision making. You will
be acting in the role of a vice president for
Conwood, Inc., a real estate development corporation.
Several decisions nee^. to be made regarding the
funding of various re. .4 estate projects in the
southeast region -- such decisions will be your
responsibility in this exercise.
You have recently been provided information
describing two office projects located in major
metropolitan areas of the southeast. The following
portfolio contains the information on each of the
projects. You should examine this information
and make a decision to fund one project as a
commercial development in your region.
All of the participants were given portfolios that included information
in the form of a financial pro forma detailing the development costs, the
expected leasing rates, the projected cash flows and the net present values
for two commercial real estate development projects, one in Birmingham and
the other in Jacksonville. A financial analysis was also included that
indicated the net present value of each project. In addition, reports from
the company's contracting specialist, market analyst and policy specialist
were included in the portfolio. These portfolios were modeled after those
used by two real estate management firms in the Southeastern U.S.
The information provided in the portfolios was constructed to be similar
for each of the projects across a number of dimensions including the net
present value, the expected demand for office space, the projected
availability of building materials and the "fit" of the project with current
corporate strategy. Similar information for each project was provided in
order to prevent a systematic choice of one project over the other from
occurring and to increase the degree of commitment that the decision makers
felt regarding their choices.
The case materials stated that the corporate policy required the vice
presidents to annually allocate a proportionate amount of the development
cost of the project based on the construction horizon of the selected
project. Since each of the projects had a construction horizon of two
years, the allocation at the ^nitial decision time should be one-half of the
total development costs for the project that was chosen for funding. At the
beginning of the next fiscal year, (i.e. the second experimental session)
the project would be reviewed and a decision regarding the next annual
allocation would be made at that time.
Within the contracting specialist's report, two levels of foreseeability
of the setback were manipulated. For those subjects in the foreseeable
condition, the contracting specialist's report stated:
The report on union activities indicates some
unrest within the masons' union A lack of resolution
regarding the use of prefabricated construction
techniques may create conflict when the current masons'
contract expires. This contract will expire during
the construction horizon for this project.
The not foreseeable condition contained no mention of labor problems.
Individuals were instructed to review the materials for each project and
to select one project to be funded. After selecting a project and
allocating funds to cover the development costs for the first construction
period, individuals were asked to write a prospectus detailing their reasons
for project selection, and were then asked to assign themselves a code name
so that their materials could be returned to them at the second session.
Self -assigned codes were used to retain anonymity.
At the second experimental session, held one week after the first one,
individuals were told that a strike (setback) had occured. The reasons for
the strike were manipulated in order to create two levels of dif fusibility .
In the high dif fusibility condition, subjects were told that a strike had
occurred. In the low dif fusibility condition, subjects were told that "they
(i.e., the subject) had approved the prefabrication of masonry slabs, off-
site, by non-union workers as a cost saving measure." They were further
told that "although this procrdure may have violated the union contract, the
step had been taken as a cost saving measure in the hope that the union
stewards would not realize it." The strike, in the non-exonerating
condition, resulted from the allocator's gamble on cost savings.
Persistence was explicitly manipulated in order to control the possible
10
effects of variation in subject's perceptions of the extent to which the
setback would continue. In the resolved condition, subjects were told that
the labor dispute had been sucessfully resolved. In the persistent
condition, they were told that although the employees were back to work,
unrest continues and a contract had not yet been signed.
The diffusibility and persistence manipulations were fully crossed with
the forseeability manipulation conducted during the first experimental -
session. The result was a complete 2 (high / low foreseeability) by 2 (high
/ low diffusibility) by 2 (resolved / persistent) experimental design.
After receiving the setback information, subjects were asked to review
the project portfolio, to make a second allocation of development costs to
the project (AMOUNT) and to decide on the amount of money that they wished
to allocate to an overage account that would be set aside for the project
(OVERAGE). Each folder also contained a memo informing subjects that a
potential buyer existed for the office building. The memo instructed
subjects to indicate their priority to sell the project on a four point
scale ranging from 1 - "low priority- -continue project" to 4 = "high
priority-- consider accepting a price covering development costs to date"
(PRTYSEL) .
After completing the case materials, subjects completed a questionnaire
containing the 5 manipulation check items displayed in the Appendix. The
first two of these items evaluated the extent to which the subjects felt
that the setback was foreseer. Ae and that they felt responsible for it. The
next two items evaluated the extent to which an outside party was
responsible for or should be blamed for the setback. The last item
evaluated the subjects' expectations that the setback would persist.
Subjects were thanked for their participation, debriefed regarding the
11
true purpose of the study and given a written feedback summary of the
experiment after all the data were collected.
RESULTS
Manipulation Checks
Analyses of variance (ANOVA's) were used to evaluate the effect of all
three experimental manipulations on the five manipulation check items. -
These results are summarized in Table 1. The manipulation of foreseeability
produced significant effects on item 1 which evaluated personal
responsibility for the setback (F 1,59 - 8.10, p < .01) and item 2 which
evaluated the foreseeability of the setback (F 1,59 = 9.17, p < .01). The
means of both items were greater in the highly foreseeable condition. The
manipulation of setback persistence produced a significant effect on item 5
which evaluated persistence (F 1,59 - 5.18, p < .05). As anticipated,
subjects in the persistent setback condition reported greater persistence
than those in the resolved condition. The manipulation of dif fusibility
produced significant effects on item 1 which evaluated personal
responsibility (F 1,59 = 25.15, p < .001), item 2 which evaluated setback
foreseeability (F 1,59 =- 13.2, p <.001), item 3 which evaluated the
responsibility of an outside party for the setback (F 1,59 = 5.83, p < .05)
and item 4 which evaluated the extent to which an outside party should be
blamed for the setback (F 1,59 = 18.07, p < .001). In the high
dif fusibility condition, subj -cts felt less personal responsibility,
perceived the setback to be less foreseeable, and attributed greater
responsibility and blame to a third party compared to the low diffusibility
condition. The ANOVA's did not reveal any other significant main effects or
interactions on the manipulation checks.
12
The contrast between the effects of foreseeability and dif fusibility on
the manipulation checks is important. As in previous studies (i.e., Staw
and Ross, 1978; Conlon and Wolf, 1980), forseeability affected the degree
to which subjects felt responsible for the setback. It did not affect the
perception of third party responsibility. In contrast, the diffusibility
manipulation both reduced the perception of personal responsibility and
increased the perception of third party responsibility for the setback.- The
unique effect of diffusion, then, was to shift responsibility away from the
subject and toward an outside party.
Dependent Measures
There were three dependent variables: AMOUNT, OVERAGE and PRTYSEL. In
the introduction, we suggested that the distribution of AMOUNT may be bi-
modal with most allocations being either the normative amount, or zero. On
examination of the frequency distribution of this measure, we found that 62
of the 67 subjects allocated exactly the projected budget requirement, 2
subjects allocated zero dollars and 3 subjects allocated amounts greater
than the projections. This result was consistent with our concerns about
the experimental demand that would be induced when projected costs were
provided to subjects. Further examination showed that the five subjects
allocating an amount other than the budget were not concentrated in any
particular experimental condition and that the two subjects who allocated
zero also placed zero in the overage account. Because the variance in
allocations was small and ess ntially random with respect to the
manipulations, we excluded the AMOUNT measure from further analyses and
tested our hypothesis with the remaining 2 indicators of commitment:
OVERAGE and PRTYSEL.
The correlation between OVERAGE and PRTYSEL was -.21 ( n - 67,
i
13
p < .05). Both the sign and significance of this relationship was
appropriate and support the notion that the two measures each evaluate
commitment. It is not suprising, however, given the differences between the
measures that we noted in the introduction, that the relationship is not
stronger. Because of the significant correlation between the two variables
and the unequal cell frequencies, a Multivariate Analysis of Variance
(MANOVA) estimating unique sums of squares was conducted to test the
hypothesis .
The means and standard deviations of the dependent variables are
presented in Table 2. The MANOVA produced two significant multivariate
effects. The strongest of these, a dif fusibility by foreseeability
interaction (Multivariate F - 5.21, df = 2,58; p < .01) was predicted by the
hypothesis. The second, a main effect for persistence (Multivariate F =
4.05, df = 2,58; p < .05) was not specified a priori . These multivariate
effects were further interpreted by examining the univariate effects on each
dependent variable. The univariate effects for OVERAGE and for PRTYSEL are
displayed in Tables 3 and 4 respectively.
It was hypothesized that diffusion of blame would interact with
foreseeability in the form outlined in Figure 1. Consistent with the
hypothesis, significant dif fusibility by foreseeability interactions were
obtained on both OVERAGE (F - 4 . 80 , df = 1 , 59 ; p < . 05) and PRTYSEL (F =
7.49, df - 1,59; p < .001). These interactions are diagrammed in Figures 2
and 3 respectively. The 95% '-mfidence intervals on the means indicated
that diffusibility affected OVERAGE in the foreseeable setback condition,
but not in the non-foreseeable condition. In the foreseeable (i.e., high
responsibility) condition, subjects allocated more to the OVERAGE account
following setbacks for which blame could not be diffused. The interaction
14
on PRTYSEL followed a similar pattern except that diffusibility produced
significant differences in both of the foreseeability conditions. When
foreseeability was high, the desire to sell was significantly greater
following the diffusible setback than following the setback that could not
be diffused. This pattern was reversed in the non-foreseeable condition
where subjects expressed a significantly greater desire to sell projects
following setbacks for which blame could not be diffused. Considered
together, the results for both OVERAGE and PRTYSEL indicate that when an
allocator is responsible for a setback, diffusibility of blame to an outside
party promotes a preference for withdrawal, whereas low diffusibility
promotes continuation in the project.
A significant univariate effect for persistence was indicated on
PRTYSEL. As would be expected, subjects indicated a higher priority to sell
the project when the labor contract had not been signed (mean = 2.36). When
the labor dispute had been successfully resolved, subjects were not as
willing to abandon the project (mean = 1.99).
A significant univariate effect for diffusibility was also found on
OVERAGE, meaning that subjects allocated more resources when responsibility
for the setback could not be diffused to the labor union.
DISCUSSION
The results of this study indicate that the behavioral manifestations
of commitment to a course of action following a setback depend not only on
the extent to which the alloc-cor feels responsible for the setback, but
also on the extent to which an outside party may be held responsible.
Although the foreseeability manipulation was sufficient to induce feelings
of responsibility in the allocators, it was not sufficient to explain how
Lb
allocators behaviorally cope with responsibility for failure. When
blame for the strike could be diffused to members of the contracting union,
withdrawal from the project was a viable option for responsible allocators.
When the blame for the strike clearly rested on the allocator, the tendency
was to continue involvement in the project.
Alternative Explanations
The design of this study ruled out several alternative explanations for
the results. Because persistence of the setback was explicitly manipulated
rather than controlled, the nature of the setback was not confounded with
expectations about its continuation. Although we found that the
manipulation of persistence had a significant effect on PRTYSEL, persistence
did not interact with either the dif fusibility or foreseeability
manipulations. It is reasonable to conclude that the effects of the latter
manipulations occurred independently of subjects' perceptions of risk or
expectations that the setback would continue.
Secondly, the interaction of forseeability and dif fusibility could not
be attributed to the joint effect of those factors on the perceived
responsibility of the allocator for the setback. As in the World Bank
scenario, the forseeability manipulation -affected levels of perceived
personal responsibility for the setback, but did not affect the perceived
extent to which a third party could be blamed. The dif fusibility
manipulation lowered the perc/ived level of personal responsibility and also
increased the perceived level of third party blame. Diffusability and
foreseeability did not interact to affect perceptions of responsibility.
The joint main effects of diffusion and foreseeability on personal
reponsibility could be used to explain the results on the commitment
16
variables if main effects were hypothesized and obtained, but the obtained
interaction cannot be explained by a simple responsibility effect. The
results suggest that the ability to diffuse blame to a third party, in this
case the union, has an effect on commitment that is independent and
different from that of personal responsibility and is not contingent on a
lowering of perceived personal responsibility.
Finally, the results obtained in this study were generally consistent
across two different indicators of commitment. Specifically, we observed
that in the foreseeable, diffusible condition, allocations to the overage
account were low and the desire to sell was high. In contrast, when
setbacks were foreseeable and blame could not be diffused, the allocation to
overage was high and the desire to sell was low. If these results had been
obtained on a single dependent variable, or if the pattern of results had
differed between the two variables, the result may have been attributed to
factors unique to allocations or sales. The convergence of the results
across the variables enhances the interpretation of these results as a
commitment phenomenon.
Implications
Several aspects of this study have implications for future research on
the escalation of commitment. The major implication concerns the manner in
which researchers construe decision making following setbacks. Staw (1976)
originally proposed that escaiaCion occurred as an attempt to rationalize
past behavior in the face of failure, a process generally referred to as
self justification. Later, Staw (1981) expanded this explanation to include
the possibility of justification to others, and a desire to appear publicly
17
consistent. The present results suggest that the behavioral manifestations
of justification (e.g., escalation, continuation or withdrawal) may depend
on the way in which the setback can be explained by the allocator. For
example, when a third party may be blamed for the setback, withdrawal can be
used to symbolize for self or others that a third party was involved in the
setback, and can be rationally justified by the allocator as a possible
means to control the future behavior of the third party. In the absence of
a third party, withdrawal may signify resignation or an admission of blame,
but continuation may help define the setback as a learning experience or as
a short term cost in a long term strategy of success. Characteristics of
the setback, such as dif fusibility of blame, may have an important effect
on the rationalization process, and more research on this issue is needed.
A second implication concerns the design of future studies of
escalation and commitment in resource allocation contexts. One feature
which explicitly differentiates the present methodology from those of
previous studies was the provision of complete financial information. There
are at least two advantages of providing such information in studies of
escalation and commitment. First, such information provides a stronger test
of "rationalizing" effects. To the extent that rationalization is described
as antithesis to "economically rational" decision behavior (Staw, 1981;
Bazerman, 1986) , studies of rationalization processes should provide
subjects with opportunities for rational choice. Studies that do not
clearly define the economical!-- rational pathway for subjects may be
criticized for creating a demand for rationalization or for failing to
distinguish between the rational and rationalizing options available to
subjects. In the present study, the economically rational signal (i.e., net
present value) favored continuation. The apparent desire of allocators to
13
withdraw in the forseeable, diffusible condition is a more impressive
demonstration of rationalizing when viewed in the context of that signal
than it would be if the economically rational path was not specified.
The second advantage of normatively relevant information concerns its effect
on internal validity. When rational paths are not specified, subjects may
infer them. To the extent that such inferences are affected by the
experimental manipulations meant to produce rationalization (e.g.
forseeability) , internal validity may be threatened. For example, if
persistence of the setback had not been explicitly manipulated and assessed
in this study, it is possible that foreseeability or dif fusibility may have
affected the allocators' expectations about persistence and rationally
affected allocations to OVERAGE. The chance of such effects occurring is
reduced by anticipating and providing a rational course of action against
which actual behavior may be compared.
The study also suggests a practical implication. The results of
studies utilizing the World Bank case (Staw and Ross, 1978; Conlon and Wolf,
1980) showed that responsibility and its effect on the need to justify a
prior behavior could promote withdrawal from a course of action as well as
escalation. The present study showed that third party involvement partially
explains the type of rationalizing strategy that allocators will use. It is
noteworthy that such strategies may appear quite rational to the parties
involved. Conlon and Wolf (1983) documented a corporate example that
parallels the results of this --tudy. In that case, a decision was made to
discontinue a new product tha;:, if kept, was projected to have an above
average profit margin for the firm. The product was discontinued because of
the persistent failure of the firm to meet development deadlines which,
because of contractual obligations to a client firm, were embarrasing to top
Li
management. At the time the decision was made to discontinue the product,
development was virtually complete and most of the "start up costs" had been
paid. The ceremony of discontinuation was a top management meeting during
which the most affected managers presented a profit and loss statement for
the product to date. The managers blamed the negative cash flow on poor
decisions by engineers who had been hired specifically to work on the new
product and who would be dismissed if the product was dropped. The
presentation ignored the known interest of potential future clients in the
product. To an uninformed observer, the process would have appeared
economically rational. Privately, the involved managers admitted their
intent to detach themselves from a product which evoked bad memories and
some embarrassment. Hence, the advice to managers on how to avoid
escalation (cf., Rubin, 1980; Brockner and Rubin, 1985) may apply equally
well to contexts of unproductive escalation and unproductive withdrawal.
20
REFERENCES
Bazerman, M.H. Judgment in Managerial Decision Making. New York: Wiley,
1986.
Bazerman, M. H. , Giuliano, T. and Appleman, A. Escalation of commitment in
individual and group decision making. Organizational Behavior and
Human Performance, 33, 1984, 141-152.
Brockner, J. and Rubin, J. Z. Entrapment in escalating conflicts : A social
psychological analysis . New York: Springer-Verlag, 1985.
Brockner, J., Nathansen, S. Friend, A., Harbeck, J., Samuelson, C. , Houser,
R. , Bazerman, M. H. , and Rubin, J. Z. The role of modeling processes
in the "knee deep in the big muddy" phenomenon. Organizational
Behavior and Human Performance , 33 , 1984, 77-99.
Conlon, E.J., and Wolf, G. The moderating effects of strategy, visibility
and involvement on allocation behavior: An extension of Staw's
escalation paradigm. Organizational Behavior and Human Performance .
26, 1980,172-192.
Conlon, E. J. and Wolf, G. The archetecture of a course of action: Case
studies and their implications. Paper presented at the Academy of
Management Meetings, Dallas, TX. , 1983.
Davis, M. and Bobko , P. Contextual effects on escalation processes in
public sector decision making. Organizational Behavior and Human
Decision Processes. 37, 1986, 121-138.
Fox, F.V. , and Staw, B.M. The trapped administrator: Effects of job
insecurity and policy resistance upon commitment to a course of action.
Administrative Science Quarterly, 24, 1979, 449-471.
Leatherwood, M. L. and Conlon, E. J. The impact of prospectively relevant
information on persistence in a project following setbacks. Working
Paper, College of Business Administration, University of Iowa, Iowa
City, IA 52242, 1986.
Lindblom, C. E. The science of muddling through. Public Administration
Review, 19, 1959, 79-88.
McCain, B. E. Continuing investment under conditions of failure: A
laboratory study of the limits to escalation. Journal of Applied
Psychology, 71, 1986, 2£ "•-284.
Northcraft, G. and Neale, M. Opportunity costs and the framing of resource
allocation decisions. Organizational Behavior and Human Decision
Processes. 37, 1986, 348-356.
Northcraft, G. and Wolf, G. Dollars, Sense and Sunk Costs: A Life Cycle
Model of Resource Allocation Decisions. The Academy of Management
Review, 9, 1984, 225-234.
i
21
Rubin, J. Z. Experimental research on third party intervention in conflict:
Toward some generalizations. Psychological Bulletin, 87, 1980,
379-391.
Rubin, J. Z. and Brockner, J. Factors affecting entrapment in waiting
situations: The Rosencrantz and Guildenstern effect. Journal of
Personality and Social Psychology, 3>I. 1975, 1054-1063.
Staw, B.M. Knee-deep in the big muddy: A study of escalating commitment to
a chosen course of action. Organizational Behavior and Human
Performance, 1976, 16, 27-44.
Staw, B. M. The escalation of commitment to a course of action. Academy of
Management Review, 6, 1981, 577-587.
Staw, B.M., and Fox, F.V. Escalation: Some determinants of commitment to a
previously chosen course of action. Human Relations , 30 . 1977, 431-
450.
Staw, B.M. and Ross, J. Commitment to a policy decision: A multi-
theoretical perspective. Administrative Science Quarterly, 23 . 1978,
40-64.
22
Table 1
Cell means for manipulation checks
CONDITION
RESPOTH
RSPSTBK BLAME
STBKFOR STBKPER
DIFFUSABLE
NOT PERSISTENT
NOT FORESEEABLE
3.87
1.62
3.75
2.37
2.75
FORESEEABLE
3.78
3.33
3.55
3.88
3.00
PERSISTENT
NOT FORESEEABLE
4.37
2.25
4.50
2.87
3.37
FORESEEABLE
4.20
2.70
4.20
3.60
3.90
NOT DIFFUSABLE
NOT PERSISTENT
NOT FORESEEABLE
3.00
3.85
2.28
3.43
3.28
FORESEEABLE
3.00
4.33
2.56
4.22
2.11
PERSISTENT
NOT FORESEEABLE
4.00
3.44
2.89
4.33
3.22
FORESEEABLE
3.57
4.14
2.57
4.57
3.10
23
Table 2
Cell means and standard deviations on AMOUNT and PRTYSEL
Condition Frequency OVERAGE PRTYSEL
DIFFUSA3LE
NOT PERSISTENT
NOT FORESEEABLE 8 1.10 1.75
(.65) (.71)
FORESEEABLE 9 .57 2.33
(.51) (.71)
PERSISTENT
NOT FORESEEABLE 8 1.06 2.13
(.82) (.35)
FORESEEABLE 10 .77 2.50
(-48) (.53)
NOT DIFFUSABLE
NOT PERSISTENT
NOT FORESEEABLE 7 1.09 2.00
(.19) (.63)
FORESEEABLE 9 1.11 1.89
(.48) (.06)
PERSISTENT
NOT FORESEEABLE 9 1.05 2.67
(.58) (.50)
FORESEEABLE 7 1.57 2.14
(1.06) (.69)
24
Table 3
Analysis of Variance on OVERAGE
Source
SS
CONSTANT
71732499
DIFFUSABILITY(A)
1810352
PERSISTENCE(B)
361133
FORES EE(C)
86180
A X B
66818
A X C
1900937
B X C
573878
A X B X C
64241
WITHIN
23381890
DF
1
1
1
1
1
1
1
1
59
MS
71732499
1810352
4
.57
361133
0
.91
86180
0
.21
66818
0
,17
1900937
4
.80
573878
1
,44
64241
0
.16
396303
**
p < .05
p < .01
25
Source
SS
CONSTANT
312.67
DIFFUSABILITY(A)
0.00
PERSISTENCE(B)
2.21
FORESEE (C)
0.11
A X B
0.15
A X C
2.62
B X C
0.40
A X B X C
0.04
WITHIN
20.62
* p < .05
** p < .01
Table 4
Analysis of Variance on PRTYSEL
DF MS
1 312.67
1 0.00 0.00
1 2.21 6.31*
1 0.11 0.31
1 0.15 0.42
1 2.62 7.49**
1 0.40 1.14
1 0.04 0.12
59 0.35
26
Figure 1
An illustration of the hypothesized interaction
low
diff usability
extent
of
commitment
high
dif fusibility
not
foreseeable
foreseeable
2 ^
Figure 2
Diagram of the obtained interaction on OVERAGE.
1.3
1.2
1.1
1.0
overage
(millions) .9
.8
.7
.6
.5
low
dif fusibility
high
dif fusibility
not
foreseeable
foreseeable
28
Figure 3
Diagram of the obtained interaction on PRTYSEL.
selling
priority
(PRTYSEL)
1 -
high
dif fusibility
low
dif fusibility
not
foreseeable
foreseeable
29
APPENDIX
Manipulation check items.
The following items were included in the post-experimental questionnaire as
manipulation checks. The anchors for the six-point scales are indicated in
parentheses following each item.
1. How much responsibility do you feel for the setback to the project?
(None - 1, A Great Deal - 6)
2. The setback to the originally chosen project was foreseeable.
(Strongly Disagree -1, Strongly Agree =■ 6)
3. How much responsibility do you feel another party had for the setback to
the project? (None - 1, A Great Deal = 6)
4. How much blame for the setback should be given to a third party?
(None - 1, A Great Deal - 6)
5. The setback to the initially chosen project will probably persist.
(Strongly Disagree -1 , Strongly Agree = 6)
30
HECKMAN IXI
BINDERY INC. |§|
JUN95
B^.To-P^ ^MANCHESTER,