The psychology of project termination
This paper presents recent research on the topic of escalation of commitment to a failing course of action. The popular press frequently reports on large projects, often government-funded projects that overrun by as much as two to three times the original budget. This raises the question as to why many of these projects are still receiving funding.
This research specifically looks at the behavior of decision makers who deal with failing projects. Within this context, this research investigates the effect that optimism has on the behavior of decision makers.
An experiment was used to test the decision preference of participants with a project management background, and their motivation for making a particular choice.
The results from the research show strong support that optimism bias is a major contributor to escalation of commitment in the selection and motivation of projects. The results also show support for the Critical Point Theory.
Escalation of Commitment (EoC) occurs when a decision maker receives information that an investment of resources will not provide the expected return, but despite this, the decision maker continues committing resources to the failing course.
Research into the phenomenon of EoC was first performed by Staw (1976), and many papers have since been written on the subject. Despite the fact that we have a much better understanding about what leads to EoC, the situation has not improved much and many organizations still pursue failing projects.
In an attempt to explain EoC, a number of reasons or determinants have been suggested. These determinants can be classified into five groups, that is, project, psychological, social, organizational, and contextual.
Meyer (2013a) identified 34 determinants from research papers since 1978. Not all of the identified determinants received equal coverage in research, and three of these determinants have been researched in considerably more detail than the others, e.g., sunk cost effect, self-justification, and project completion (Sleesman, Conlon, McNamara, & Miles, 2012).
Sunk Cost Effect
Sunk cost in the context of projects is defined as “a cost that has already been incurred and which should not be considered in making a new investment decision” (Amos, 2007, p. A.11).
Arkes and Blumer (1985) present a number of experiments to show the propensity of subjects to favor investments with a higher sunk cost compared to investments with a lower sunk cost. This finding is supported by Whyte (1986, 1993), who also notes the important role of decision framing in escalation.
Sunk cost effect research presents strong evidence that there is a correlation between the willingness to spend more on a project (that will most likely not make any money for the organization) and an increase in the sunk cost (Coleman, 2010; Cunha & Caldieraro, 2009).
Staw (1976) investigates the effect of personal responsibility on EoC in situations where a decision maker had a choice between alternatives and had the opportunity, at a later stage, to make further decisions to influence the situation, that is, the decision maker can allocate additional resources or withhold resources. He argues that a decision maker may cognitively distort the negative consequences of a prior decision, instead of changing his or her behavior, to make it appear to be a rational and favorable decision.
The decision maker should, however, have committed to a situation that cannot be easily changed (Brehm & Cohen, 1962) and feel that he or she would be held personally responsible for the negative consequences of the decision (Carlsmith & Freedman, 1968; Cooper, 1971).
In an analysis of research results of EoC prior to 1992, Brockner (1992) concludes that self-justification is one of the primary reasons for EoC. Brockner (1992) posits that, in past research, there is a strong argument that decision makers will primarily consider their own prior decisions when deciding to invest more resources in a failing project.
Conlon and Garland (1993) argue that project completion and sunk cost are cofounded, and that prior research may have overemphasized the effect of sunk cost as a reason for EoC. Research by Garland, Sandefur, and Rogers (1990) shows that the opposite of the sunk cost effect may occur, that is, less is invested as the incremental cost of the project increases. On a typical project, the amount of money spent is correlated to the progression of the available time to complete the project. This relationship is not linear, and graphing the incremental project cost over time typically takes the form of an S-curve (PMI, 2013).
A decision maker would usually be aware of the time progress of the project when decisions are made, and that the end of the project is approaching. The approaching end of the project could entrap the decision maker, and may lead to goal substitution, whereby decision makers shift their focus away from the goal the project has to achieve in terms of its deliverables, to the goal of just completing the project (Ting, 2011).
Jensen, Conlon, Humphrey, and Moon (2011) posit that decision makers will increasingly conceal negative information about a project as the project approaches completion.
The Effect of Optimism Bias
An area that has received little attention in previous EoC research is optimism bias. Humans have a tendency to be overly positive when we evaluate ourselves, and overstating our own ability to master or control situations appears to be part of our normal thought processes (Lovallo & Kahneman, 2003; Taylor & Brown, 1988, 1994).
Some argue that this optimism bias is a method of survival to ensure that we take risks based on the prospect of a better outcome, even if there is very little or no substance that the anticipated outcome is possible. Our optimism bias is difficult to control, and even if decision makers are aware of this bias, they are very unlikely to control or alter their behavior (Juliusson, 2006; Tali Sharot et al., 2012).
Sharot (2012) found that optimism is seated in the inferior frontal gyrus (IFG) part of the brain. Interfering with the physical functioning of the IFG through transcranial magnetic stimulation decreases optimism bias, suggesting that optimism bias is a biological attribute of humans.
The effect of optimistic behavior on projects is that a decision maker may select a project based on an overestimation of the value that the project will add to the organization or an overestimation of their own ability to influence the outcome of the project (Bazerman & Samuelson, 1983). Once the project is selected, the decision maker may be optimistic that he or she can control the outcome of a failing project in a positive way, even if all the supporting evidence shows otherwise, due to the illusion of control.
Critical Point Theory
Meyer (2013b) defined the Critical Point Theory (CPT) as a framework for explaining how post-project optimism bias affects the decision maker and why EoC takes place.
The CPT is summarized as follows: A decision maker in an escalating project situation will construct a valuation of the benefit of the project based on his or her perception of a set of measurable and immeasurable gains that will be produced by the project. In the same way, a decision maker will construct a valuation of the investment in the project based on his or her perception of the measurable and immeasurable resources that will be committed to the project.
Each gain is assigned a subjective weight, and the sum of the gain/weight combinations is the total perceived benefit of the project. In the same way each committed resource is assigned a subjective weight to construct a total investment value. As long as the perceived benefit is greater than the perceived investment of the project, the decision maker will continue to escalate commitment to the project. When the decision maker's perceived benefit is equal to or less than the perceived investment, the critical point is reached, and the decision maker will stop the project.
The business benefit from the project is defined as:
B = business benefit,
w = relative weight of the contribution of gain g to the benefit,
g = the gain contributed to the benefit, and
m = total number of weight-gain pairs.
The investment in the project is defined as:
I = investment,
w = relative weight of the contribution of resource r to the investment,
r = any resource that contributes to the investment, and
n = total number of weight-gain pairs.
It is important to note that:
- Gains (g) and resources (r) can be expressed in any unit to reflect all of he contributors to the investment and the gain. A project that has no financial benefit could, for example, be expressed in terms of the increase in employee morale.
- Decision makers deal with two benefit and two investment values. The first is the planned investment (Ip) and planned benefit (Bp). These values are calculated before the project starts, and the initiation of the project is often based on the magnitude of the difference between these values, that is, Bp should exceed Ip for a feasible project. The second values are the estimated investment (Ipt) and the estimated business benefit (Bpt) at any point in time during the project, but before the project end.
Given the past research on optimism bias, it is conceivable that project decision makers may be overly optimistic about the future outcome of a project. This optimism can take two forms. The first form is in-project optimism, where decision makers believe that, even if the project is in a bad state now, it will still be delivered within the original time and cost parameters. This type of optimism may be the result of prior experiences of project success in adverse conditions (Staw & Ross, 1978), or is simply due to irrational optimism (Tali Sharot et al., 2012). From this we define the first hypothesis:
H1. Project decision makers will continue to invest in a failing project, based on their optimism that the project will eventually recover and deliver, within the original time and cost targets.
The second form is optimism about the business benefit from the project. Decision makers believe that the return on investment from the project will be greater than what was calculated in the project's business case. No specific reference to post-project optimism bias was found in the literature review of this research. From this we define the second hypothesis:
H2. Project decision makers will continue to invest in a failing project, based on their optimism that the project will give better returns than what was found in the project business case.
Assessment Design and Decision Problem
A five-group post-test-only randomized experiment jointly varied an invested amount and project progress over five intervals. The five intervals were correlated to the amount of money spent in comparison to the original budget. The intervals were 20%, 40%, 60%, 80%, and 110%.
The experiment was designed to follow previous experiments that tested for EoC, where a low project completion and low cost scenario is compared to a high completion and high cost scenario (Conlon & Garland, 1993; Garland & Conlon, 1998; Jensen et al., 2011; Moon et al., 2003; Tversky & Kahneman, 1981). This design, therefore, allows for comparison to previous research in this area.
Presented with the scenario below (20% values), participants had to indicate whether or not they would continue with a project. The values for the 40%, 60%, 80%, and 110% scenarios are shown in brackets. The financial information provided in the scenario is sufficient for participants to calculate the ROI prior to the project, and at the specific assessment point in time of the scenario. Participants were told that they could do any calculation to arrive at their answer, and that they could use pocket calculators for this purpose.
You are a manager in your organization and a member of a committee responsible for recommending and selecting projects.
Your company is conducting a project, which you recommended with a planned cost of $25 million and a planned duration of 15 months.
Before the project was started, the estimated return on investment (ROI) was 30%.
The project is now at the end of month 2 (5, 7, 10, 15), and the project manager has reported that the project is likely to take four months longer (i.e., 19 months) to complete.
The actual cost to date for the project is $5M ($10M, $15M, $20M, $27.5M), and the project manager estimates that the project will cost $8.5M more than originally planned (i.e., $33.5M).
At a project review meeting, a decision must be made whether to continue with this project or not. If the project is stopped now, approximately $2.5M ($5M, $7.5M, $10M, $14M) of the investment can be salvaged in the form of equipment and material. The remainder is a sunk cost attributed to labor and consulting fees.
If you stop the project, the unused money and salvaged equipment and material will be redirected toward other new or existing projects in the company.
a. Stop the project.
b. Continue with the project and approve the additional $8.5M.
After making a choice for the continuation or termination of the project, participants had to indicate to what extent they agreed with the statements in Exhibit 1, in support of their continuation/termination decision. Participants could choose from the following Likert scale of agreement:
|Strongly Disagree||Disagree||Disagree Somewhat||Undecided||Agree Somewhat||Agree||Strongly Agree|
Any score greater than 0 for a statement, therefore, shows some measure of support.
Exhibit 1—Percentage of Respondents Who Decided to Continue with the Project per Scenario
There were 482 responses received, of which 308 (64%) chose to continue with the project, and 174 (36%) chose to stop the project. The participants were specifically targeted as individuals who had some involvement in projects and project management. Approximately 51% of the participants were post-graduate students on part-time project management courses at two South African universities, and the remaining 49% completed the survey online.
The respondents were from 38 countries and 28 different industries: 85% of the participants had bachelor degrees or higher degrees; 63% indicated that they worked in organizations with more than 1000 employees; 56% indicated that the average project duration was 6 to 24 months, and 74% reported that the value of projects in their organization was below $100M.
Exhibit 2 shows the percentage of respondents who chose to continue with the project.
Exhibit 2 – Percentage of respondents who decided to continue with the project per scenario
The statistical tests used in both hypotheses are t-tests. Before the t-test results are accepted, three checks were done on the data.
- Independence of Observations: Each case in the data set represents a single respondent, that is, the data do not show repeated measurements of the same respondents.
- Outliers: Boxplots of the data revealed no outliers.
- Test for Data Normality: Visual inspection of Q-Q plots of the data shows that the data are normally distributed.
Test of Hypothesis 1
To test H1, it must be shown that there is statistically significant support for MS10. To test the significance, the following test is performed:
- Null Hypothesis: H0: A decision maker will assess that managerial efforts have no effect on one's decision to stop or continue a project.
- Statistical Test: A t-test is done to determine if there is a difference between the observed assessment of the decision maker's own managerial efforts, and the expected assessment. The test is done across all the project scenarios, and the test value is set at 0.0 since a value ≤ 0.0 would indicate that H0 cannot be rejected.
- Significance Level: Let α = 0.05 (95% confidence), with N = 308 (respondents who chose to continue with the project). The critical test value for the t ratio is 1.64 at 95% confidence (Larsen & Marx, 2012, p. 701).
- Test Results: MS10 is positively supported in S20 to S110 (t (307) = 10.68, p < 0.001) for project continuation (see Exhibit 3). The choice of a decision maker to continue with the failing project, and the associated support for MS10, supports the notion that decision makers would continue with a failing project while citing the possibility of recovering the project to meet the planned schedule and budget, as a reason for doing so. H0 is, therefore, rejected for H1, and the alternative hypothesis is accepted.
Exhibit 3 – t-Test Results for Statement MS10
Test of Hypothesis 2
To test H2, it must be shown that there is statistically significant support for MS5, MS7, and MS8. These statements indicate the respondent's belief that the project's completion will have benefits for which there is no supporting information, that is, based on optimism. To test the significance, the following tests are performed:
Null hypothesis: H0: Decision makers will assess the estimated business benefit from the project to be the same as that shown by the information currently available to them through the project's business case.
Statistical Test: A t-test is performed on the data to determine if there is a difference between the observed assessment of the project's business benefits and the expected assessment.
The expected assessment for MS5, MS7, and MS8 is that a decision maker will show neutral or negative support for the motivation statements. The test is done across all the project scenarios, and the test value is set at 0.0 since a value ≤ 0.0 would indicate that H0 cannot be rejected.
Significance level: let α = 0.05 (95% confidence), with N = 308 (respondents who continue with the project). The critical test value for the t ratio is 1.64 at 95% confidence (Larsen & Marx, 2012, p. 701).
Test Results: The results of the t-test for each scenario are shown in the exhibits below.
Exhibit 4 – t-Test Results for Statement MS5
Exhibit 5 – t-Test Results for Statement MS7
Exhibit 6 – t-Test Results for Statement MS8
MS5 is positively supported in S20 to S110 (t(307) = 13.80, p < 0.001) for project continuation; Exhibit 4. This result supports the notion that decision makers feel that the project will provide greater business benefits after project completion. There is no information available to support this idea; this support must, therefore, be subconsciously constructed by the decision maker.
MS7 is positively supported in S20 to S110 (t(307) = 27.27, p < 0.001) for project continuation; Exhibit 5. These results support the notion that decision makers feel that there will be benefits, other than those directly attributable to the project, from the project. There is no evidence in the project description that supports this idea, and this support must, therefore, be subconsciously constructed by the decision maker.
MS8 is positively supported in S20 to S110 (t(307) = 20.34, p < 0.001) for project continuation; Exhibit 6. These results support the notion that decision makers feel that not all the benefits from the project will be lost. This assertion is clearly incorrect, as a simple calculation, with the data available in the project description, shows that the project will lose money. Therefore, the decision makers must be overly optimistic about the performance of the project.
The results of this study show strong support for the notion that decision makers will continue with a project that is clearly failing, because they are optimistic about the future of the project. The study also shows that decision makers will consider multiple determinants when assessing the continuation of a failing project.
The results from this study support the Critical Point Theory (Meyer, 2013b) and show that decision makers’ perceptions of the value of a product changes over the life of a project, and that these perceptions are mainly due to optimism bias.
The implication of this research for managers is that decision makers in control of project portfolios may continue to invest in failing projects, simply because they are optimistic about the project improving or giving better business results than planned. When left unchecked, failing projects may consume considerable resources that could have been used for other projects. A number of such cases have been studied, such as the Shoreham nuclear power plant (Ross & Staw, 1993), world's fair Expo86 (Ross & Staw, 1986), the Chicago Deep Tunnel Project (Staw & Ross, 1987), and Barings Bank (Jensen et al., 2011).
A number of strategies have been proposed to defuse escalation situations. Strategies with a high level of success in studied escalation situations are briefly described as follows:
Staw and Ross (1987) suggest replacing decision makers in the decision-making situation, to break escalation behavior. Removing a decision maker would remove the internal justification psychological determinants, but would not remove the project determinants, and would provide partial remedy for social and structural determinants (Keil, 1995; Ross & Staw, 1993).
Support for Failure
Staw and Ross (1987) suggest that organizations provide limited support for managers who have had failed projects. Managers are not at risk of losing their jobs, but could be restricted from running larger projects for a period of time, following a project failure. This would reduce the high stakes a decision maker faces for making mistakes, which in turn would reduce commitment.
Bringing Phase-out Costs Forward
Staw and Ross (1987) suggest that presenting the cost of stopping a project to decision makers before a project is started will create consciousness of the withdrawal cost. It would, furthermore, be beneficial if withdrawal costs are shown at significant project review points throughout the life of the project.
Early and Frequent Risk Assessment
Assessment of project risks should start at the early stages of a project, and should be continued through the life of the project. The project team should specifically question the continuation of the project at frequent intervals (Keil, 1995).
Minimum Goal Setting
Decision makers were asked to set minimum target levels of project performance which, if not achieved, would lead to a change in policy and possible withdrawal. Minimum goal setting resulted in the biggest reduction in EoC of the factors tested by Simonson and Staw (1992).
The study only looked at the effect of financial benefits on the decision makers. The effect of optimism bias about nonmonetary benefits, such as company image, sustainable business growth, employee well-being, etc., was not tested.
Group decision making was not considered. The experiment was aimed at individual decision makers who may be part of a project selection committee.
A control group was not set up to test the level of support for the given statements for a project that was not in trouble. This research tests for the relative change in behavior between the low and high positions. A control group could be set up in future experiments, when the sample group is large enough or when a different data collection strategy is used.
Future research in this area should consider the previously mentioned study limitations in the design of the experiment, for example, testing for other forms of benefits.
Experiments, of the type used in this research, become complex when more parameters are tested. A more feasible approach for decision making (and specifically EoC in projects) may be computer simulation games.
This research shows strong support that decision makers are optimistic that the product will somehow perform better than what they assessed it to be before the project started. It also differentiates clearly between in-project and post-project optimism bias.
It is, however, possible for managers to put measures in place to give early warning signs of escalating situations and to reduce its effect.
This research makes three important contributions to the existing body of knowledge of EoC:
- It confirms that decision makers are influenced by multiple determinants when assessing the continuation of a failing project;
- In-project and post-project optimism bias are defined as significant contributors to EoC, which should be considered in future studies in this area; and
- This research shows support for the Critical Point Theory and confirms that decision makers have a strong belief that a project will have greater benefits than originally anticipated, with no supporting evidence.
Amos, S. (Ed.). (2007). Skills and Knowledge of Cost Engineering (5th ed.). Morgantown, WV: AACE International.
Arkes, H. R., & Blumer, C. (1985). The Psychology of Sunk Cost. Organizational Behavior and Human Decision Making, 35, 124–140.
Bazerman, M. H., & Samuelson, W. F. (1983). I Won the Auction But Don't Want the Prize. Journal Of Conflict Resolution, 27(4), 618–634.
Brehm, J. W., & Cohen, A. R. (1962). Explorations in Cognitive Dissonance. Hoboken, NJ: John Wiley & Sons Inc.
Brockner, J. (1992). The Escalation of Commitment to a Failing Course of Action: Toward Theoretical Progress. The Academy of Management Review, 17(1), 39–61.
Carlsmith, J. M., & Freedman, J. L. (Eds.). (1968). Bad Decisions and Dissonance: Nobody's Perfect. Chicago, IL: Rand McNally.
Coleman, M. D. (2010). Sunk Cost and Commitment to Medical Treatment. Current Psychology, 29(2), 121–134.
Conlon, D. E., & Garland, H. (1993). The Role of Project Completion Information in Resource Allocation Decisions. The Academy of Management Journal, 36(2), 402–413.
Cooper, J. (1971). Personal Responsibility and Dissonance: The Role of Foreseen Consequences. Journal of Personality and Social Psychology, 18(3).
Cunha, M., & Caldieraro, F. (2009). Sunk-Cost Effects on Purely Behavioral Investments. Cognitive Science, 33(1), 105–113.
Garland, H., & Conlon, D. E. (1998). Too Close to Quit: The Role of Project Completion in Maintaining Commitment. Journal of Applied Social Psychology, 28(22), 2025–2048.
Garland, H., Sandefur, C. A., & Rogers, A. C. (1990). De-escalation of Commitment in Oil Exploration: When Sunk Costs and Negative Feedback Coincide. Journal of Applied Psychology, 75(6), 721–727.
Jensen, J. M., Conlon, D. E., Humphrey, S. E., & Moon, H. (2011). The Consequences of Completion: How Level of Completion Influences Information Concealment by Decision Makers. Journal of Applied Social Psychology, 41(2), 401–428.
Juliusson, A. (2006). Cognition and Neurosciences—Optimism as Modifier of Escalation of Commitment. Scandinavian Journal of Psychology, 47(5), 345–348.
Keil, M. (1995). Pulling the Plug: Software Project Management and the Problem of Project Escalation. MIS Quarterly, 19(4), 421–447.
Larsen, R. J., & Marx, M. L. (2012). An Introduction to Mathematical Statistics and its Applications (5th ed.). Boston, MA: Prentice Hall.
Lovallo, D., & Kahneman, D. (2003, July). Delusions of Success: How Optimism Undermines Executives’ Decisions. Harvard Business Review.
Meyer, W. G. (2013a, June 2013). The Effect of Optimism Bias on the Decision to Terminate Failing Projects. Paper presented at the IRNOP XI, Oslo, Norway.
Meyer, W. G. (2013b, May 2013). The Critical Point—The Psychology of Project Termination. Paper presented at the PMI Global Congress 2013, Instanbul, Turkey.
Moon, H., Conlon, D. E., Humphrey, S. E., Quigley, N., Devers, C. E., & Nowakowski, J. M. (2003). Group Decision Process and Incrementalism in Organizational Decision Making. Organizational Behavior and Human Decision Processes, 92, 67–79.
PMI. (2013). A Guide to the Project Management Body of Knowledge (PMBOK® Guides—Fifth Edition. Newtown Square, PA: Project Management Institute.
Ross, J., & Staw, B. M. (1986). Expo 86: An Escalation Prototype. Administrative Science Quarterly, 31(2), 274–297.
Ross, J., & Staw, B. M. (1993). Organizational Escalation and Exit: The Case of the Shoreham Nuclear Power Plant. The Academy of Management Journal, 36, 701–732.
Simonson, I., & Staw, B. M. (1992). Deescalation Strategies: A Comparison of Techniques for Reducing Commitment to Losing Courses of Action. Journal of Applied Psychology, 77(4), 419–426.
Sleesman, D. J., Conlon, D. E., McNamara, G., & Miles, & J. E. (2012). Cleaning up the Big Muddy: A Meta-analytic Review of the Determinants of Escalation of Commitment. Academy of Management Journal, 55(3), 541–562.
Staw, B. M. (1976). Knee-deep in the Big Muddy: A Study of Escalating Commitment to a Chosen Course of Action. Organizational Behavior and Human Resourses, 16, 27–44.
Staw, B. M., & Ross, J. (1978). Commitment to a Policy Decision: A Multitheoretical Perspective. Administrative Science Quarterly, 23, 40–64.
Staw, B. M., & Ross, J. (1987). Behavior in Escalation Situations: Antecedents, Prototypes and Solutions. Research in Organizational Behavior, 9, 39–78.
Sharot, T., Kanai, R., Marston, D., Korn,C. W., Rees, G., & Dolan, R. J. (2012). Selectively Altering Belief Formation in the Human Brain. Proceedings of the National Academy of Sciences of the United States, 109(42), 17058–17062.
Taylor, S. E., & Brown, J. D. (1988). Illusion and Well-Being: A Social Psychological Perspective on Mental Health. Psychological Bulletin, 103(2), 193–210.
Taylor, S. E., & Brown, J. D. (1994). Positive Illusions and Weil-Being Revisited: Separating Fact from Fiction. Psychological Bulletin, 116(1), 21–27.
Ting, H. (2011). The Effects of Goal Distance and Value in Escalation of Commitment. Current Psychology, 30(1), 93–104.
Tversky, A., & Kahneman, D. (1981). Framing of Decisions and the Psychology of Choice. Science, 211(4481), 453–458.
Whyte, G. (1986). Escalating Commitment to a Course of Action: A Reinterpretation. The Academy of Management Review, 11(2), 311–321.
Whyte, G. (1993). Escalating Commitment in Individual and Group Decision Making: A Prospect Theory Approach. Organizational Behavior and Human Decision Processes, 54(3), 430–455.
© 2013, Werner G. Meyer
Originally published as a part of 2013 PMI Global Congress Proceedings – New Orleans, USA