Sunday, April 25, 2004
Back from Chicago
I was in Chicago for a few days for the Midwest Political Science Assoc meeting. Nice conference. And nice to not be watching the news for a few days. Then I got a cold, which I've just recovered from. So it's been a while since my last update.
*****
Thoughts on what I learned at Midwest panels: Comparative political economists like to study interactive models---e.g., models where the effect of some variable on an outcome depends on the context---but we have yet to fully come to grips with how to discriminate among interactive models. I am a big user of these techniques, and I include myself in this complaint. For any problem we wish to study, there are lots of possible model specifications (each considering a different style of interaction, or conjuctual causation), and it's often hard to choose on theoretical or empirical grounds which, if any, is right. If we had enough data---that is, enough countries and time periods that are comparable in all respects but the interactive variables in question---we could probably get somewhere by assessing relative goodness of fit. But with the dozens of countries to hand, this isn't usually persuasive. And it is hard to definitely theorize about these relationships. So what can we do? I see two broad possibilities:
1. Go micro
2. Emphasis cross-validation
The "go micro" option means testing the causal links in the chain, empirically grounding theoretical models so that we are more confident of our specifications when we turn to macro questions. This is hard, but probably the best option. Option 2, cross-validation, is something we can do more easily: show that our models work on "reserved" cases, essentially an exercise in prediction.
We can do either or both, or other things I haven't listed. But we'll have to do more if we want to convince outsiders of our findings. (FYI, going linear additive is not on the list: if we have good reason to think context matters, but we're not sure how, pretending it doesn't just gives answers we know are probably wrong.)
*****
Other thoughts on the news from the last two weeks:
Since I've been away, the meat-grinder that is Iraq ground on, Bush sided with Israel and Sharon on Israeli settlements in the West Bank, and Bush finally started making noises to the UN. Plus, Colin Powell appears to be aching to get his tell-all (or, "it wasn't my fault") book off his chest, and seems to be going to Bob Woodward to leak it out. But he tells us nothing we didn't know already: Bush wanted war with Iraq, started planning after 9/11, Powell opposed it internally but played along for the audience, and essentially sold us all out. And now he wants to look foresighted? Good luck.
Tom Friedman has a great all-purpose line from Paul Romer, the Stanford economist: "A crisis is a terrible thing to waste." It would be nice to think we had the kind of informed, clever, thoughful leadership that could seize such moments---but I think this bunch of know-nothings needs to first learn the first law of holes (when you're in one, stop digging). How they handle the Falluja stand-off may be the key test.
On the other hand, I am pleased Bush is finally going to the UN for help, and I am crossing my fingers that rumors Chalabi is headed for the chop are true. It may be too late to salvage the situation, but we have no choice but to try. On the other hand, I can't help but sympathize with anyone---from Halliburton's employees to the Spanish military---that wants to get out of the trap Bush set for them.
*****
Thoughts on what I learned at Midwest panels: Comparative political economists like to study interactive models---e.g., models where the effect of some variable on an outcome depends on the context---but we have yet to fully come to grips with how to discriminate among interactive models. I am a big user of these techniques, and I include myself in this complaint. For any problem we wish to study, there are lots of possible model specifications (each considering a different style of interaction, or conjuctual causation), and it's often hard to choose on theoretical or empirical grounds which, if any, is right. If we had enough data---that is, enough countries and time periods that are comparable in all respects but the interactive variables in question---we could probably get somewhere by assessing relative goodness of fit. But with the dozens of countries to hand, this isn't usually persuasive. And it is hard to definitely theorize about these relationships. So what can we do? I see two broad possibilities:
1. Go micro
2. Emphasis cross-validation
The "go micro" option means testing the causal links in the chain, empirically grounding theoretical models so that we are more confident of our specifications when we turn to macro questions. This is hard, but probably the best option. Option 2, cross-validation, is something we can do more easily: show that our models work on "reserved" cases, essentially an exercise in prediction.
We can do either or both, or other things I haven't listed. But we'll have to do more if we want to convince outsiders of our findings. (FYI, going linear additive is not on the list: if we have good reason to think context matters, but we're not sure how, pretending it doesn't just gives answers we know are probably wrong.)
*****
Other thoughts on the news from the last two weeks:
Since I've been away, the meat-grinder that is Iraq ground on, Bush sided with Israel and Sharon on Israeli settlements in the West Bank, and Bush finally started making noises to the UN. Plus, Colin Powell appears to be aching to get his tell-all (or, "it wasn't my fault") book off his chest, and seems to be going to Bob Woodward to leak it out. But he tells us nothing we didn't know already: Bush wanted war with Iraq, started planning after 9/11, Powell opposed it internally but played along for the audience, and essentially sold us all out. And now he wants to look foresighted? Good luck.
Tom Friedman has a great all-purpose line from Paul Romer, the Stanford economist: "A crisis is a terrible thing to waste." It would be nice to think we had the kind of informed, clever, thoughful leadership that could seize such moments---but I think this bunch of know-nothings needs to first learn the first law of holes (when you're in one, stop digging). How they handle the Falluja stand-off may be the key test.
On the other hand, I am pleased Bush is finally going to the UN for help, and I am crossing my fingers that rumors Chalabi is headed for the chop are true. It may be too late to salvage the situation, but we have no choice but to try. On the other hand, I can't help but sympathize with anyone---from Halliburton's employees to the Spanish military---that wants to get out of the trap Bush set for them.