Some thoughts on the relationship between political science, journalism and political consultancy spurred by “this response”:http://blogs.princeton.edu/mccarty/2009/05/a-reponse-to-kk.html by Nolan McCarty to Keith Krehbiel. Note that I make _no guarantee whatsoever_ that these thoughts are original – just that it _may_ be generally useful to explicitly lay out the underlying assumptions of debate, even when lots of people implicitly know what those assumptions are.
First – what political scientists are most interested in is the explication and explanation of the causal mechanisms underlying broad patterns of political outcomes. This comes out clearly in the argument between McCarty and Krehbiel. As McCarty notes, the problem with _ex ante_ predictions is that the good ones may turn out (in a single instance) to be wrong, and the bad ones (in a single instance) to be right. This is a claim that Krehbiel almost certainly agrees with – political science’s ability to make predictions will at best be messy and imperfect, and in any event are only a sidebar to the main story that political science wants to tell, which is a story about why things happen the way they do. Obviously, a really good causal theory would both explain _and_ predict accurately nearly all of the time, but the most we can reasonably aspire to are more-or-less-good theories which will be somewhat helpful in providing predictions, but will be noisy and, often enough, wrong. The theoretical proof of the pudding is usually (albeit not always) in the broader explanation of patterns, rather than the specific ability to predict or not predict event _x._
Second – what journalists are usually interested in from political scientists _is_ prediction of single events or of the behaviour of individual actors. Smarter journalists _are_ interested in explanation too, but it usually takes second place to the imperatives of a story that is usually driven by the news cycle, and by specific events unfolding in that cycle. Arlen Specter’s defection is a perfect example of this. While journalists’ main question is _how does this affect whether health care etc will be passed_, McCarty and Krehbiel are more fundamentally interested (I think), in the question _what does Specter’s defection and subsequent behaviour mean for my underlying theories of politics?_ This creates problems for the relationship between social scientists and journalists. Accounts such as Krehbiel’s and McCarty, Poole and Rosenthal’s _do_ have significant predictive insights for how individual actors will behave under changed circumstances. If they didn’t, then they would probably not be very useful. But their applications for punditry and prognostication are a side benefit – what they are supposed to do is explain _patterns of outcomes over time_ (where you can expect at least some of the idiosyncracies of individual actors etc to wash out). Thus, the dialogue between social scientists and journalists is always going to be a little tricky, although it can be a useful one (and is one that is, imo, improving substantially).
Third, political operatives want answers to a very different question – what is going to help my guy (no gender specificity intended) get elected? Here, political scientists often don’t have very much to say. Many of our best models of electoral outcomes suggest that strategy, rhetoric and the other tools that are part and parcel of the politician’s toolkit, are so much froth, and that the underlying determinants are factors such as economic growth etc that are hard to jimmy (although by no means impossible). This is both uncongenial to political operatives under the Upton Sinclair principle, and not very useful. There is, of course, one significant area of cross-over, which is Green et al.’s work on the relative efficacy of different forms of voter contact etc. This does provide useful information to political operatives, which is being taken up. And perhaps the single most valuable lesson that political science has to offer here can be summed up in two words – Randomized and Experiments. But large chunks of the rest of political science are _structurally irrelevant_ to political operatives – the kinds of knowledge that they seek to generate are not the kinds of knowledge that help you win elections. Nor is this necessarily a bad thing (as long, in my purely personal opinion, as that knowledge is pragmatically useful or interesting in some other way).
This is, in some ways, the distillation of what I have learned from writing posts for this blog over the last year, and reading others’ posts and comments. As noted at the beginning, this is not exactly revelatory – even so, I figure I am thinking more clearly about these different goals, and their relationships, than I was before I started blogging here, and thought that it was worth sharing (others’ mileage may vary, of course).