Model Behavior

Bruce Bueno de Mesquita’s article ("Recipe for Failure," November 2009) and The Predictioneer’s Game, the book from which it is drawn, present political outcomes as stemming from the preferences of individual politicians, groups, or states. However, the choice in many situations is between particular policies and not eventual outcomes. Because Bueno de Mesquita ignores the relation between preferences and concrete policies, his model explains much less than he thinks it does.

Whether they’re deciding what the United States should do in Afghanistan or which course of action the international community should follow on global warming, leaders face choices about specific policies. Should combat forces be sent to a particular province at a particular time? Should cap-and-trade arrangements of particular sorts be implemented over a certain period of time?

The fact that particular actors may have preferences for certain outcomes means little-at least without some kind of assessment as to the probability that a given policy will result in a specific outcome. Because, as we have discussed elsewhere, those assessments are often anything but exact, in practice, policymaking revolves around which policy options are already in place, most easily adjusted, or readily available: in other words, around organizational capabilities and not actors’ preferences.

Bueno de Mesquita’s model appears to work because all the heavy lifting about policy alternatives has already been done. He simply feeds the existing policy options into the model as a starting point. Hence, the decisions he models as being taken by actors with specific outcome preferences are tantamount to cutting the ribbon after a construction project has finished.

Bruce Bueno de Mesquita replies:

I thank Stephen Majeski and David Sylvan — longtime critics of my efforts to model foreign affairs — for taking the time to try to think about the policy process from a modeler’s point of view. Unfortunately, I have, apparently, not been clear enough for them to have properly understood my approach and how it is used in practice. Their ribbon-cutting metaphor misses the mark.

As I explain in The Predictioneer’s Game and elsewhere, one of the essential methods of predicting outcomes is to focus on alternative what-if scenarios and alternative ways of framing issues. The modeling effort provides the opportunity to look at alternative strategies and alternative policies, sorting out analytically which are likely to produce better or worse outcomes and then attempting to frame the issues around the preferred policies and approaches of the decision-makers whose interests one is trying to analyze.

Organizational capabilities, of course, constrain choices. But that is one of the reasons why my model’s inputs examine what stakeholders say they want rather than what might be their true preference. After all, the positions they stake out on issues are strategically chosen, taking into account what they think is feasible given the personal and organizational constraints they face.

The Predictioneer’s Game opens with an example of just such institutional and organizational constraints shaping policy positions when it discusses the differences between the actions of Leopold II in Belgium and in the Congo Free State. Perhaps when Majeski and Sylvan have a chance to read the full book they will see that there is no incompatibility between my approach and their perspective except for the formalization of the logic in my models.