The financial and economic turmoil of the years 2007–10 has led to considerable regret among financial and economic policymakers about bad policy decisions at earlier dates. A worthwhile exercise in economic analysis is a careful delineation of the net economic cost of an earlier bad policy decision. Such an analysis is conceptually difficult because it requires a baseline case against which to compare observed economic outcomes. Comparing the actual outcome to that from the ex post best possible policy decisions at every juncture gives an unrealistically high benchmark, because it compares the actual outcome to that from policy decisions requiring perfect foresight by policymakers. Also, rational evaluation requires that all gains and losses subsequent to a policy decision be included. It is incorrect to evaluate an earlier past decision based on present and future impacts, since any intermediate impacts between the past decision date and current evaluation date must also be considered. This paper suggests a theoretically simple and well-defined procedure for analysing the ex post economic cost of an isolated policy decision. We suggest the comparison of actual economic outcomes with those that would have emanated from a counterfactual but real-world feasible alternative decision. We do this by identifying a specific policy choice available at the time of the decision which was ‘almost’ chosen. We call this almost-but-not-chosen policy the sliding doors choice. The cumulative economic welfare difference between relevant economic outcomes under this counterfactual choice and under the actual choice over all post-decision periods is our measure of the ex post net economic cost of the actual policy decision. Although we do not follow their particular methodology, we invoke Leeper and Zha’s (2003) ‘modest policy intervention’ theory, in which small variations in policy do not alter agents’ rational expectations, thereby circumventing the Lucas (1976) critique of counterfactual policy analysis. We argue that this restrictive evaluation procedure can be illuminating within certain narrow circumstances.