8.1 SOME MINI "WAR STORIES"
8.1.1 Don't Take My Men Away!
As our first example, we focus on a spatial and temporal reallocation of personnel within an urban service system. Any time that allocation levels are changed, certain groups or individuals are likely to "gain" while others are likely to "lose," at least as measured on a perceptual basis. Feedback from these groups and individuals may suggest modifications of the objectives, performance measures, and/or constraints under which the analysis is carried out. This happened in the late 1960s in the New York City Police Department when queueing analysis showed that police patrol cars could be "traded" during certain hours among police precincts in southern Manhattan to reduce queueing delays. While the total number of patrol cars in the set of precincts concerned would remain constant, Erlang's formulas (see Section 4.5) showed that, because of differences in the 24-hour demand patterns among the precincts, substantial reductions in queueing delays could be achieved. The trading was to be implemented by sending one or two of precinct A's cars to precinct B at appointed hours, when the latter precinct had unusually high demand. At other hours, when precinct A's demand was relatively high, reciprocal transfers would occur. In a pilot run, carefully planned and monitored, queueing delays were reduced by about the amounts anticipated.
However, precinct commanders, when on duty during the "depleted hours," found the plan unacceptable. Although it was shown that queueing delays had not increased measurably during the depleted hours in their precincts, the commanders felt that they would be unable to handle a serious emergency (e.g., large fire, civil disturbance) in their precincts, given its depleted resources, without requesting assistance from other precincts. Quantified, their perceived objective appeared not to be queueing delay reduction but something close to a "minimax" strategy, minimization of the deleterious consequences of the largest conceivable event requiring police services that could occur in their precincts. However, many nonquantifiable factors also pervaded the issue, including the perceived power associated with the number of officers under one's command, the dependence associated with relying on someone else's resources in times of possible large-scale emergency, the cutting of traditional officer-supervisor relationships by having the officers work temporarily under different supervisors, the resistance to change brought about by directive from headquarters, and so on. As a consequence, this effort to vary patrol levels by time of day was abandoned.
In this case, none of the model-oriented factors in the study were problematic. However, because the model users (headquarters staff members) did not encompass all relevant decision makers-some of whom had objectives strongly conflicting with the objectives of the model users-implementation of the program as originally designed was infeasible.1 1 Subsequent analyses led to alternative, more politically acceptable reallocations. See, for example, [LEVI 78].