Case
Escalation of commitment
Escalation of commitment is persistence on the strength of prior investment when current evidence points the other way. Barry Staw described the mechanism in 1976, and Kettei answers it with abort criteria set before sunk cost builds up.
Failing projects keep going for a reason that has been measured. In Staw's 1976 experiment, each of 240 business students allocated research money to 1 of 2 divisions in a company case, and some later learned that the chosen division had declined. Students who had made the first choice personally and watched it fail put more of a fresh allocation into the failing division than any other group did.
Staw's explanation was self-justification, the need to show the original decision had been sound. An organisation turns that individual pull into policy. For the executive who approved a project, stopping it means admitting the approval was flawed. Each review that closes without a decision to stop is, in effect, a decision to continue.
The case
On 11 March 1993 the board of the London Stock Exchange stopped work on Taurus, the system meant to take British share settlement off paper. The Exchange said it had spent £75 million, and the Independent reported the next morning that the City faced losses of at least £275 million. Many financial institutions, the paper reported, had known of fundamental problems for at least 18 months.
In the project's final weeks the board asked an accountant to investigate, and his answer was another 2 to 3 years at up to double the cost so far. Once a number was on the table, the board stopped within about 10 days. The failure on the record came earlier, and the 4 accounts cited in the essay record no condition, set at the start in 1987, under which the project would end.
What a written line does
Itamar Simonson and Staw tested remedies on a similar case in 1992. The baseline group gave the losing product a mean of US$5.1 million of a fresh US$10 million. Participants who wrote down minimum sales and profit targets straight after the first choice, before any results, gave it US$3.9 million.
A thorough list of pros and cons left the figure at US$5.1 million, and telling participants each would be judged on the outcome tended to raise it. Even with targets, nearly 40% of the new money went to the product that had missed its targets. The written condition reduced the over-commitment and left much of it in place.
What Kettei asks
The remedy is the abort criteria, the conditions under which a decision is revoked. Each is defined at formation, before sunk cost accumulates, and each is binary and observable with a named owner. Once a criterion is met, the Owner declares the abort and the decision ends outright, with no revalidation.
An abort criterion for Taurus in 1987 might have been a delivery date and a cost ceiling, owned by one person. The accountant's finding would then have met a line drawn before the money went in. The Exchange might still have chosen to go on, by a new decision taken in the open.
The question to ask
Without such a line, continuing is the standing policy of a commitment, and at Taurus it ran until an outsider priced the alternative. What would have to be seen to stop, and was that written before the money went in?
Read the original
Escalation Is a Policy was published on Kettei on Substack on 28 October 2026, and is gathered in Kettei Decision Digest 4: Failures of Stopping, Part 1. The experiments are set out with the other sources at The research behind decision governance.
Sources
- 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 Performance 16(1), 27-44. Establishes the experiment and the self-justification account.
- Simonson, I. and 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. DOI 10.1037/0021-9010.77.4.419. Establishes the comparison of remedies.
- House of Commons (1993) Budget Statement, section "TAURUS" (Chancellor of the Exchequer, Norman Lamont). HC Deb 16 March 1993 vol 221 c178. The official record of the consequence for stamp duty.
- Willcock, J. (1993) City in 275m pounds computer fiasco: Stock Exchange chief resigns after Taurus shares system is scrapped, threatening 1,000 jobs. The Independent, 12 March 1993. Establishes the sums reported and the 18 months.
- Drummond, H. (1996) Escalation in Decision-Making: The Tragedy of Taurus. Oxford University Press. Cited from the publisher's record.
From the essay Escalation Is a Policy on Kettei on Substack, 28 October 2026.

Kettei
The Discipline of Governed Intent
