Reactions to "A Model Discipline"

I figured a good use of this space is to provide my thoughts on important recent books or articles relevant to research in public management.  This may sometimes be a discuss an article directly related to the study of public management.  But, I am going to start with some reactions to a provocative new book in political methodology -- A Model Discipline:  Political Science and the Logic of Representations by Clarke and Primo.




I will break my comments up in a few parts, but I wanted to start with an important theme from the book.  Clarke and Primo repeatedly emphasize that one never compares a theoretical model to data.  One always compares a theoretical model to a model of empirical data.  This serves to question the privilege sometimes given to empirical observations over abstract or formal models.

It is worth providing a little background.  Clarke and Primo are reacting to a specific movement within political science.  At one point a few years back, the American Journal of Political Science (through its editorial board) stated that they would not accept articles that included only a theoretical model without some empirical evaluation.  Basically, even a fundamental advance within game theory would not get published without some empirical assessment of that advance.

This change in editorial policy caused a lot of push back (predictably, from the community that relied most on game theory and formal models).  The support for the movement came from the EITM (Empirical Implications of Theoretical Models) movement.  The result are fairly clear battle lines that appear starkly with Clarke and Primo's argument.

Clarke and Primo's proposition that one only compares theoretical models the models of empirical data is a useful reminder.  Whenever you conduct empirical research, you are modeling the empirical data in some way.  You have models involved in any measurement strategy.  You have models in any descriptive statistics (a mean is a model of an underlying distribution).  You clearly have models when you move towards inference, prediction, or anything like that.  Clarke and Primo's statement serves to remind us that empirical data is not "pure" in a way that privileges these exercises over theoretical models.

My reaction to the argument, though, is a little different.  I agree with Clarke and Primo's insistence that everything we do involves models.  However, I think theoretical models are different than empirical models.  They both have important uses (and the next post will talk more about their description of the purposes of models).  The advantage I see to combining empirical and theoretical models (as in EITM work) is that each approach to research is prone to different errors.  It is not that empirical work is better than theoretical work.  Instead, combining good empirical work with good theoretical work is is protective against errors in the accumulation of knowledge.  Rather than debating one approach to modeling is superior to the other, I would like to emphasize the advantage of combining the approaches.

In the next post, I will look at Clarke and Primo's discussion of the goals of models as a basis for model evaluation.