One of the things that amuses me about this debate is that, with few exceptions--Gaddis being one of the more notable ones--historians tend to be very methodologically sloppy. They do good archival work but don't feel any compunction about picking the facts that fit their theory and telling the story they want to tell.

I agree with everything said in this post. I think a great number of qualitative researchers will be offended too by Adesnik's antropological view on what the discipline ought to be like. By the way, most departments still have a language requirement or allow substitution between languages and statistics. Political science is actually quite a diverse discipline and will remain so.

Another nice example of "large N-work" that gives different insights than deeply immersing oneself in a single case is that of james fearon and davide laitin on civil wars. While they don't argue that ethnic, economic and religious grievances are unimportant, they point out that such grievances exist in a great many countries and in only a few of them civil wars erupt. This leads them to an analysis of the conditions under which such grievances are likely to lead to insurgency, a point missed by most case studies. for a nice summary see: http://news-service.stanford.edu/news/september25/civilwar-925.html

"Is Adesnik really suggesting that people like Edward Said can enlighten us about the region?"

I think the general point though has very little to do with Said, though he makes for a convenient punching bag. For a lot of the things political science studies there are no datasets, and in many cases learning the language and/or culture of a foreign country is the best way to do some research. I don't think the argument is that this leads "automatically" to greater understanding. (Much like having a great p-value doesn't make for more rigorous or scientific results.)

Since I don't know anything about the PS field, for me the thing that popped out about David's essay was this:

"For example in support of the assertion that suicide attacks are not connected to religion....The [Tamil] Tigers' behavior only reinforces the belief that suicide bombing is a product of ideological extremism..."

But the idea that it's extremist ideology, not religion per se, was Pape's whole point, wasn't it? David actually seems to be agreeing with him, while saying he's not.

I'm sure you didn't know this at the time, and I have no love for Edward Said's work or beliefs at all, but it is unfortunate that you criticize him by name on right when he has just died.

Hey! Don't misquote me. I wasn't agreeing with Adesnik. I just think that the discipline should not exclude qualitative research. APSR hasn't published a qualitative paper in ages. This is leading to some resentment against quant-jocks. But not from me!

Laura, Pape's paper is, for all intents and purposes, a qualitative paper. He presents data, but all the inference is qualitative (eyeballing the numbers and drawing conclusions based on that, rather than using any regressions or chi-squares or anything remotely quantitative).

Speaking of the August APSR, my copy seems to have disappeared... I wonder what I did with it? Grr.

BTW, Josh Chafetz now has a post up at OxBlog dissenting in part and agreeing in part.

Glad to see people are interested in this esoteric topic. Not glad that no one agrees with me!

In any event, my response is now up on OxBlog.

Although I would tend to agree with Dave on the abuse of statistics, I'm rather more in favor of persons knowing enough about the use of them in order that they may attempt to abuse them in the first place.

That is to say: at least with a (sic)language - albeit mathematical - we can disagree on the same terms. Otherwise Poli-Sci might as well go by Poli-div(inity).

How about this: I'll support your argument to have language requirements replace static at the university - if you support my movement to cram a statistics requirement up the soft-gray mass of the nation's sixth-graders?

David is RIGHT on the money to send a warning shot to political "scientists" attempting to quantify their ART.

I am a 20 year practitioner in finance and an author of several academic journal articles. The academic portion of finance has been hijacked for the last 50 years by quants and they are in the process of ruining the practicing portion with their ideas of causality and precision. (they have already ruined the economics profession - now a laughing stock in the marketplace)

The problem for the quants in the social sciences is that their models etc must predict HUMAN BEHAVIOUR. It is not enough to explain what HAPPENED (we know this already)or relationships (we know this intuitively), you must be able to predict what WILL happen. In short, time varying parameters (and variables) and shifting distributions over time. Lack of autocorrelation yada yada. It can't be done. Give it up. It's crap.

Go back to your world of intellectual insecurity. You are smarter than you think.

Quoth Ken:

"The problem for the quants in the social sciences is that their models etc must predict HUMAN BEHAVIOUR. It is not enough to explain what HAPPENED (we know this already)or relationships (we know this intuitively), you must be able to predict what WILL happen."

Ken, I think that you misplaced the emphasis in your post; the emphasis should have been on the word PREDICT, and not on the words following (human behavior).
If this (prediction) is the criterion of "scientificness", then meteorology, AND aeronautic engineering are NOT sciences, since both of them confront real-world problems that are analytically intractable, and therefore, neither field is able to deterministically predict outcomes. (Incidentally, they get around this problem by simulation, which, in my opinion, is the direction in which political science should be moving.)

The more I do analysis of large n dta sets, the less I like it. Measurement issues and model specification all too often give researchers crucial leeway. I'm not talking John Lott here, instead I'm talking about the decisions about which measure of poverty to control for, etc. THe term of art in the academy is "publication bias" but it can have an ideological bent as well.

And all too often we fetishize our measurements. Which in retrospect is an interesting choice of words, but there it is. For example, maternal education is frequently used as a proxy for parental background. It does leave out a lot of other relevant info, but if you have maternal education data, you've covered yourself. Speaking of education, we use any test data we can find as an outcome measure, even if the test isn't based on curriculum, has a ridiculous psychometric margin of error etc. And don't get me started on the legislative studies practice of using interest group ratings to control for legislative ideology such that models explain the votes of members of congress based on how they voted in the past. Predictive, yes. Adding to our understanding? Not likely.

Can quantitative analysis, maximum likelhood estimation, etc add to our understanding? Yes. Of course. But here, I'm reminded of John Lott again. It can increase your understanding even more than a gun can keep your home safe. But it is still subject to similar misfirings.

The more I do analysis of large n dta sets, the less I like it. Measurement issues and model specification all too often give researchers crucial leeway. I'm not talking John Lott here, instead I'm talking about the decisions about which measure of poverty to control for, etc. THe term of art in the academy is "publication bias" but it can have an ideological bent as well.

And all too often we fetishize our measurements. Which in retrospect is an interesting choice of words, but there it is. For example, maternal education is frequently used as a proxy for parental background. It does leave out a lot of other relevant info, but if you have maternal education data, you've covered yourself. Speaking of education, we use any test data we can find as an outcome measure, even if the test isn't based on curriculum, has a ridiculous psychometric margin of error etc. And don't get me started on the legislative studies practice of using interest group ratings to control for legislative ideology such that models explain the votes of members of congress based on how they voted in the past. Predictive, yes. Adding to our understanding? Not likely.

Can quantitative analysis, maximum likelhood estimation, etc add to our understanding? Yes. Of course. But here, I'm reminded of John Lott again. It can increase your understanding even more than a gun can keep your home safe. But it is still subject to similar misfirings.