Cheers!

Dan:

I like where your methodology is going (quantitative is golden). I see a few problems you need to take into account:

1. Weighing: Currently, if I am reading it correctly, all your DPs are weighed equality. However, the importance of the different DPs are not equal. The way you have it now, your data management is actually probably causing more of an effect as the variables themselves.

2. You are using all domestic data, however “neighborhoods matter”. Try including a binary of whether any countries bordering the state have had a war in the last five years. There are lots of good research out there on this.

3. (and this is really KingDaddy's point), Your model has to address both the succeeding states inside the gap and the failing states outside the gap.

I think if you took into account just the first two points the map would look very different, and Botswana would be removed from the gap.

I'll check this post later today.

P.S. Calling the poster's comments a “hit and run” is slightly over-dramatic.

Yeah. You are.

Jay,

Agreed.

Chris,

Thanks for your comments, regarding my test of the core/gap model [1]!

I liked your clean organization into four subpoints, so I'll answer those individually

1) “Currently, if I am reading it correctly, all your DPs are weighed equality. “

I was interested that Barnett created a categorization model based on service response days, and then provided metrics that supported such a scheme. I'm sure there are better methods of testing and operationalization the variables, but I was impressed not only did the core/gap model pass face analysis, but that it produced falsifiable hypohesis that not only generally supported the model, but generated further hypotheses (such as my contention that the “Gap” may just be the set of African and Islamic states)

2) “Try including a binary of whether any countries bordering the state have had a war in the last five years.”

What dataset/threshold of violence would you use. (A serious question. I've been focusing on psychological studies recently, but it would be interesting to see if that data is availabl.)

3) “Your model has to address both the succeeding states inside the gap and the failing states outside the gap.”

Not really. Think of core/gap as a condition, but as a factor: great at explaining between-group differences, not great at explaining within-group differences.

4) “Try including a variable on income inequality (Gini Coefficient is good). That might take care of some of the outliers.”

A superficial glance [2] doesn't show a matchup between core/gap — my guess being that income inequality can be an outcome of economic growth more than economic wealth.

Finally)

“P.S. Calling the poster's comments a “hit and run” is slightly over-dramatic.”

Would “a dictation disguised as a dialogue waste of my time?” have been less dramatic?

[1] http://www.tdaxp.com/archive/2006/05/08/redefining-the-gap-1-prologue.html
[2] http://en.wikipedia.org/wiki/Image:World_Map_Gini_coefficient.svg