Occasionally, when you have really talented graduate students, you are looking for a book that would challenge them to think deeply about their research. This is such a book. James Mahoney and Dietrich Rueschemeyer have put together an excellent volume that takes up some of the most challenging problems regarding questions about research design, how to think about causality in the social sciences, and how to connect questions about ontology to research methods.

In Comparative Historical Analysis in the Social Sciences, a number of seasoned veterans reflect on research they and their compatriots have been doing over the last 25 years or so. It reads like a number of interesting and detailed reports from various academic battlefields. The picture one gets is a meeting where staff and frontline officers reflect on what has been accomplished in the latest battles. What they have been doing, why they did it, and why their strategy and approach is preferable to others are the questions they try to answer.

Editors Mahoney and Rueschemeyer give a good introduction to the special characteristics that have distinguished the comparative historical research approach. It is interesting to note that this is not a research program marked by a common theory. Nor is it, unless in the broadest sense of the word, united by a common methodology. When it comes to theory, you will find scholars who give preference to the importance of institutions, to cultural traits, or to state elites, to name a few. Regarding methods, researchers in this field are unashamedly opportunistic. They will use anything (statistical large-n analysis, archival data, elite interviews, surveys) to solve their research problems. What seem to unite researchers in this field are three things. One is that their research is driven by the formulation of puzzles that are created by comparing historical cases. A typical question would be “Why do a couple of otherwise similar countries have very different types of welfare states?” Or why did country X and country Y, that are very different, experience similar social revolutions?

The second distinguishing quality is taking causality as connected to temporality very seriously. Such outcomes as revolutions, national differences in systems of industrial relations, or variation in the outcome of major public policies can be explained. Here, a clear line is drawn against postmodernist and interpretivist approaches.

The third characteristic is a healthy skepticism toward all-embracing universal theories, such as the rational choice theory or the theory about modernization. The argument for this is simple, yet important. The variation we can see around the world is simply too great to be explained by any such “theory of everything.” Clear examples would be variation in corruption, or variation in union density. In game-theoretical terminology, researchers in this field are interested in explaining why agents, faced with similar problems, end up in very different equilibria. To take one example, in France, less than 10% of all workers are union members, while in Sweden over 85% join unions. It is certainly impossible to explain this by referring to any form of pure individual rationality. What we need is a historical explanation of why the “payoff” of being a union member has become so different in various countries. In a similar vein, a theory positing that both Palermo and Oslo can be explained by referring to agents' cultural understanding of the state is not very helpful. What we need are historical explanations of why such different cultural understandings of the state have come to exist, and how the causal mechanism that reproduce these different cultural understandings operate.

One criticism that has been launched against this research is that it is inferior to large-n statistical approaches in producing universal explanations. In his chapter, Jack Goldstone gives an excellent rejoinder to this criticism. One of his arguments is that historical case studies should be understood as examples of Bayesian analysis, which is based on strong priors. If you, for example, have a strong theory stating that social revolutions are caused by class-based social forces, one case that shows otherwise (e.g., by pointing at the importance of state elites) can be “strikingly important.”

Paul Pierson and Kathleen Thelen both present chapters showing different ways of handling time and sequences of events in historical explanations. Pierson gives a fresh account of how to view “threshold effects” and of how and when agents can suddenly change their expectations of others' likely behavior. This approach should be important in cases like the outbreak of ethnic violence in the former Yugoslavia. Thelen's chapter serves to remind us about the stickiness of many political institutions and how we can explain institutional evolution. Her example (skill formation in different systems of industrial relations) is fascinating.

In his chapter, Peter A. Hall points at the discrepancy between recent developments in our ontological understanding of how societal causality operates, and our standard repertoire of methods, especially the use of multivariate regression analysis in comparative politics. His claim that “the ontologies of comparative politics have substantially outrun its methodologies” (p. 398) is a new and important argument. His first argument is that we often assume unit homogeneity while we know that this is not the case. For example, six years of conservative rule in the 1930s are not equivalent to six years of conservative rule in the 1980s. Secondly, the development of an ontology that recognizes strong feedback mechanisms, lock-in effects, and autocorrelation between variables over time is not compatible with the idea that the world consists of variables than can be clearly distinguished by labeling them “independent” and “dependent.” Strategic interaction or institutionally induced payoffs that serve to strengthen the reproduction of that very institution are but two examples of this problem. Thirdly, we have observations that the event(s) that ultimately put(s) a system onto a specific historical “path” leading to a unique equilibrium may have occurred at “formative moments” very early in the process. Hall's point is well taken, namely, that such ultimately important variables to be found in a “distant past” are hard to capture by using the standard regression method. His main recommendation for aligning ontology and methodology in comparative politics is that analysis should be centered on the tracing of processes so that we can uncover how the causal mechanisms operate.

To summarize, researchers working in this approach have certain characteristics that are to be applauded. They do not have a “Theory” for which they only look for data they know beforehand will prove this Theory. And they have not invested so much time and energy in a certain method that they feel they have to use it whether it is appropriate or not. Instead, they are curious about how to explain variation in important political outcomes, and they are willing to use a variety of theories and methods to underpin their explanations for these outcomes. In the end, they may only accomplish tentative explanations of these outcomes, but they do produce a lot of interesting things to read.