Analytic Narratives Revisited

We welcome the animated debate raised by Analytic Narratives concerning social scientific methods and the scope of rational choice. Advocates of mathematical and rational models have long claimed they have much to tell us about situations where behavior can be quantified or where the situation under study recurs many times. However, it was thought impermissible for rational choice theories (and rational choice) to venture into the analysis of big events. Political scientists like Gary King, Robert Keohane, and Sidney Verba (1994) implicitly conceded the issue by concentrating on the problem of case selection when the number of cases is small but greater than one. We believe unique events are too important to leave aside, and we use rational choice, particularly game theory, as a means to study unique events.

A symposium on AN is a difficult exercise. The writing of analytic narratives is still in its infancy, and the topics and aims of the volume range across disciplines and over more than a millennium. The commentaries by Daniel Carpenter, Sunita Parikh, and Theda Skocpol reflect a patience and openness that we can only applaud. Overall they agree on the merits of the enterprise but debate the nature, relevance, and extensiveness of our contribution. The question that we must therefore confront is not whether to craft analytic narratives but what constitute the standards for research in this vein.

Our critics perceptively indict us for a number of misdemeanors and perhaps even a few felonies. To most of Carpenter’s, Parikh’s, and Skocpol’s charges we plead guilty with honor. Rather than responding to each of their criticisms individually we recognize that they fundamentally concern four issues: (1) Does AN actually deliver what the introduction promises? (2) Where is the narrative? (3) Where is the analytical method? (4) How do we transform an approach to problems into a research area in social science?

Moving from the Introduction to the Cases

Is analytic narrative a method or an approach? This important question remains unresolved among the authors and consequently is evaded or inconsistently treated throughout the book. The commentary authors and other critics note that there is dissonance between the claims of the introduction and the content of the chapters. The dissonance is partly a reflection of an introduction written as a manifesto versus chapters written to advance substantive knowledge and analysis. But the dissonance runs deeper than that. Each chapter establishes that rational choice models can be used to organize a narrative of important moments in history, yet none goes very far in establishing how this can be done for moments other than those the chapter considers.

Carpenter, Skocpol, and Parikh argue convincingly that it is too early to equate analytic narrative with a method. As an approach, analytic narrative is most attractive to scholars who seek to evaluate the strength of parsimonious causal mechanisms. The requirement of explicit theorizing compels scholars to make causal statements and, if the model is to be fully explicated, to identify a small number of variables.

The advantages of this approach are twofold. First, it provides the researcher with some discipline. If the model is important, the narrative should not rely overmuch on factors outside the model. Hence, an explicit theory allows the scholar to distill the narrative. Second, research of this kind can proceed by iteration. A given narrative suggests a model that when explicated ought to have implications for the structure of relationships (the institutions) within which the events occurred. Those implications force the scholar to reconsider the narrative and then to reevaluate the extent to which key elements of the narrative lie outside the proposed theory. If one must appeal too often to forces outside the model, then the theory must be rejected. Analytic narrative is therefore an inductive approach that challenges both the evidence about the event of interest and theories that structure that evidence. Considered...