Bendor and Hammond: Rethinking Allison's models

Bendor and Hammond. 1992. Rethinking Allison's models. American Political Science Review 86 (June): 301-22.

A highly critical review of Allison's 1969 article and 1971 book about decision making and the Cuban missile crisis. They attack all three of Allison's models. They accuse him of straw-manning rational choice in Model 1 and therefore discarding it prematurely. They argue that Model 2 is just plain wrong: there is no reason to expect simple rules (standard operating procedures) to yield simple outcomes. And they argue that Model 3 is not only flawed, but too complex to be useful for anything but history.

Generally, they group their many criticisms under five general heads:

1. All models begin by applying some prior set of assumptions to an analysis of the world. Allison's problem: in all 3 models, he makes mistakes in applying ideas previously expressed in other literatures. Model 1 forgets that many rational choice models seek to solve games of _uncertainty_. In Model 2, Allison concluded that standard operating procedures ("simple rules") constrain rational organizational behavior, but this was a misapplication of Simon's and March's arguments that standard procedures and divisions of labor enable boundedly rational individuals to produce fully rational outcomes. In Model 3, Allison completely neglects the role of hierarchy in bureaucratic bargaining, and he disregards the (frequently discussed) conflict between expertise and authority.
2. Allison champions the idea that we should be aware of our underlying assumptions, and of how competing models yield differing predictions, so that we can compare the relative usefulness of different models. Yet Allison has straw-manned rational choice, which prevents him from seriously comparing it to his proposed models.
3. Furthermore, Allison argues that analysts must make their assumptions explicit. Unfortunately, Allison fails to adequately take his own advice. He left his two models' assumptions "sufficiently ambiguous that it is difficult to discern the models' defining properties." As a result, Bendor and Hammond have trouble divining whether Models 2 and 3 belong in the same category or not.
4. Moreover, Allison fails to rigorously derive his conclusions from his assumptions. Bendor and Hammond find Allison's conclusion (from Model 2) that "simple rules lead to simple outcomes" completely misguided. They also complain that Model 3 "is so complicated that virtually no propositions can be rigorously derived from it at all."
5. Finally, Bendor and Hammond note that Allison's empirical evidence is useless: If predictions are not rigorously derived from assumptions, then corroborating evidence has little meaning.

Bendor and Hammond make sound arguments. They are correct in most of their criticisms of Allison. I wonder, though, how much this matters at times. For example, they attack Allison for giving too "thin" a description of rational choice. But was Allison really trying to test rational choice, or was he simply trying to illustrate the ways in which his ideas differ?