Keil and Escalation


  • Keil, M., Mixon, R., Saarinen, T., & Tuunainen, V. (1994). Understanding Runaway Information Technology Projects: Results from an International Research Program Based on Escalation Theory. Journal of Management Information Systems, 11(3), 65-85.
  • Keil, M., Tan, B.C.Y., Wei, K.K., Saarinen, T., Tuunainen, V., & Wassenaar, A. (2000). A cross-cultural study on escalation of commitment behavior in software projects. MIS Quarterly, 24(2), 299-325.

Other Sources

  • Brockner, J. 1992. The escalation of commitment to a failing course of action: toward theoretical progress. Academy of Management Review, 17(1) 39-61.
  • Festinger, L. (1957). A Theory of Cognitive Dissonance. Evanston, IL: Row, Peterson and Company.
  • Garland, H. (1990). Throwing good money after bad: the effect of sunk costs on the decision to escalate commitment to an ongoing project. Journal of Applied Psychology, 75(6), 728-731.
  • Kahneman, D., and Tvcrsky, A. (1979). Prospect theory: an analysis of decisions under risk. Econometrica, 47, 263-291.
  • Kahneman, D., and Tvcrsky, A. (1982). The psychology of preferences. Scientific American, 246, 160-173.
  • Staw, B.M., and Fox, F.V. (1977). Escalation: the determinants of commitment to a chosen course of action. Human Relations, 30(5), 431-450.
  • Whyte, G. (1986). Escalating commitment to a course of action: a reinterpretation. Academy of Management Review, 11(2), 311-321.


There is a line of research within Project Management that looks at “runaway” IT projects. This research seeks to understand why project teams continue to commit resources to failing projects. It’s referred to as “escalating commitment””

Escalating commitment has been defined as continued commitment in the face of negative information about prior resource allocations coupled with ‘uncertainty surrounding the likelihood of goal attainment’ (Brockner, 1992). Within the escalation literature, the term ‘escalating commitment’ does not necessarily imply an increasing rate of investment over time, but rather, refers to a growth in the cumulative amount of resources invested over time. Thus, escalating commitment can best be thought of as continued commitment (Keil et al., 1994, p. 80).

Colloquially, “throwing good money after bad.”

Keil et al. examine the phenomenon through a handful of theoretical frameworks: Self-Justification Theory, Prospect Theory, and Risk-taking Theory.

Self-justification theory (SJT) [emphasis mine], which is grounded in Festinger’s (1957) theory of cognitive dissonance, is based on the notion that “individuals seek to rationalize their previous behavior . . . against a perceived error in judgment” (Staw and Fox, 1977) and suggests that in the presence of negative feedback, individuals who have made a prior commitment to a project are more willing to commit additional resources than those without any prior commitment (Keil et al., 1994, p. 68)

This is consistent with Swann’s self-verification theory: We seek input confirming our self-image and filter inputs when necessary to maintain our self-image. Someone faced with inputs that show their project is failing, whether those inputs are from project data or from management opinion, would tend to distrust the negative input in favor of maintaining a self-image of competence. (Assuming, in this case, that the individual perceives themselves as competent in their role on the project team. Someone with a negative self-image, believing themselves less competent, may be quicker to accept negative feedback and stop the escalation of a doomed project.)

While this self-image as a competent team member could involve many types of identities, it is difficult to imagine it in the context of self-identification with the product under development.

Prospect theory [emphasis mine] provides a framework for understanding cognitive biases that influence human decision making under conditions of risk uncertainty (Kahneman & Tvcrsky 1979; Kahneman & Tvcrsky 1982). As such, it describes the heuristics that people use to evaluate risky choices. Prospect theory posits that individuals exhibit risk-averse or risking behavior depending on how a problem is framed. In short, prospect theory suggests that individuals will exhibit risk-seeking behavior in choosing between two negative alternatives (especially when the choice is between a sure loss – the initial loss on investment- and the possibility of a larger loss combined with a chance to return the reference point (Whyte, 1986)). Garland (1990) has invoked prospect theory to explain so-called sunk cost effect- the tendency to “throw good money after bad” (Keil et al., 1994, p. 69).

I’m no expert on prospect theory, but its independent variables seem to be the traits of choices and the independent variable a level of accepted risk. It doesn’t assert that one’s identities affect this behavior but posits the resulting behavior of any identity given certain inputs. Since “individuals will exhibit risk-seeking behavior” when the alternatives all seem negative, one might expect the framing of the choice to cancel a project as more important, generally, than who is making the choice.

Risk-taking theory [emphasis mine] suggests that risk perception and risk propensity of individuals affect their risk behavior (Sitkin and Pablo 1992). Given that the decision to continue a software project is risk-seeking behavior, risk perception and risk propensity are likely to affect decision makers’ willingness to continue a project.6 Risk perception is “a decision maker’s assessment of the risk inherent in a situation” (Sitkin and Pablo 1992). Based on our definition of risk, an event is considered risky if its outcome is uncertain and may result in a loss (Barki et al. 1993; Mellers and Chang 1994). Risk propensity is the tendency of a decision maker to take risky actions (Kogan and Wallach 1964; Sitkin and Pablo 1992) (Keil et al., 2000, p. 303).

Risk-taking theory makes the individuals the center of attention again, rather than the framing of choices. One’s risk perception and risk propensity are factors. Risk propensity might be a measure of a person identity (“I’m a risk-taker”; “I’m cautious”) or be tied to a group identity or role identity. It’s a bit harder for me to imagine it as part of a material identity standard.

In research, cultural variations in escalation behavior are examined, but only in the “national personality” sense of culture:

These factors are assessed for cross-cultural robustness using matching laboratory experiments carried out in three cultures (Finland, the Netherlands, and Singapore) (Keil et al., 2000, p. 300).

In identity terms: How have residents of each of these countries internalized the meaning found in their societies such that their risk-taking behavior might be generalizable and unique from other societies’ project team members? We don’t know which identities are being activated. We don’t know if Fins might behave differently because of their sense of what it means to be Finish or some other group, role, person, or material identity created within that national context.

Floyd et al. (1989) also posits a national character when it comes to software development. Multi-national, actually, since Floyd talks about defining traits of Scandinavian software development. If these works assert culture can affect project behavior, and we accept that our identities are the internalization of cultural meaning, these articles lend legitimacy to propositions regarding other identities affecting project behavior, including self-identification with the IT under development.

I also thought this was interesting:

Second, we decided to change the context of the case scenario from a project for internal application (limited potential payoff) to a project for external sale (high potential payoff) (Keil et al., 1994, p. 72).

This changes the cultural context of the product and the project team. It changes the nature of the technology under development, moving it closer to the cultural end of the utilitarian-cultural continuum. That may change the way project members self-identify with that technology. It also mimics the risk-reward traits of cultural goods more closely than the typical historic MIS research context.

Possibly more later.

Leave a Reply

Your email address will not be published. Required fields are marked *