1 edit
Well I just read his paper. Anyone interested can check it out by searching for "Econometrics and Presidential Elections." It was published in the Journal of Economic Perspectives (1996) Volume 10 Number 3.
I thought the paper was pretty non-technical, and I found his modest manner refreshing. He did point out several times that he was estimating based upon a very small set of observations, so that performance could be highly sensitive to changes in the data.
One thing that stuck out to me at the end of the paper was his prediction that the 1996 election would be very close. In fact, his first choice was a narrow Republican victory. Didn't Clinton win that election pretty easily over Dole?
Originally posted by telerionI don't think he justified his model too well, and I think it's rather simplistic although I have essentially no understanding of politics or economics.
Well I just read his paper. Anyone interested can check it out by searching for "Econometrics and Presidential Elections." It was published in the Journal of Economic Perspectives (1996) Volume 10 Number 3.
I thought the paper was pretty non-technical, and I found his modest manner refreshing. He did point out several times that he was estima ...[text shortened]... was a narrow Republican victory. Didn't Clinton win that election pretty easily over Dole?
Actually, I didn't read his paper until just now. The 'coolest thing' aspect is how he handled the interviewer.
Originally posted by royalchickenWell, his model is simplistic. Unfortunately with that few observations there are only so many parameters that you can introduce into the model. He believes that the few economic variables that he uses are sufficient, and his test statistics look pretty good. In econometrics we do not have the luxury of repeated experimentation. So rather than try to explain almost everything that factors into the model; we try to find some statistically significant ones that will serve. The rest, as he says, goes into the error term.
I don't think he justified his model too well, and I think it's rather simplistic although I have essentially no understanding of politics or economics.
Actually, I didn't read his paper until just now. The 'coolest thing' aspect is how he handled the interviewer.
Pardon me if you found that pedantic. 🙁
I have a few other questions for him as well. Like why did he use OLS to estimate the parameters. In order to do this you have to make some strong assumptions about the distribution of the errors, particularly homoskedasticity and no serial autocorrelation. But doesn't he assume that past realizations show up affect future? In fact, isn't that one of his central tenets, that agents are backward-looking?
To be fair to him again, this sounds more like a hobby model than a career-making one. I mean it's published in the Journal of Economic Perspectives, not exactly a "Big 5" journal in economics.
Originally posted by royalchickenHe says his model averages a 2.5% error (he may mean standard deviation, but that's not too important). I suspect he may have committed the number one sin of economists, and the reason they're shunned by natural scientists: making a model, tweaking the numbers to fit the data, and then using the same data to show that the model is good.
http://www.nytimes.com/2004/08/15/magazine/15QUESTIONS.html?ex=1094961600&en=7b737a629533714b&ei=5070
Originally posted by AcolyteI don't agree Acolyte. He's simply created a model that better fits the data. The author acknowledges that the real test is how it works post 1996. Again I think the overwhelming problem for him is that his sample is so small.
He says his model averages a 2.5% error (he may mean standard deviation, but that's not too important). I suspect he may have committed the number one sin of economists, and the reason they're shunned by natural scientists: making a model, tweaking the numbers to fit the data, and then using the same data to show that the model is good.
Standard error does not exactly mean standard deviation. Standard error is a type of standard deviation. Here's a nice site that clears up the difference. http://www2.chass.ncsu.edu/garson/pa765/normal.htm
Economists really aren't "shunned" by natural scientists. No doubt there are a some jabs that get thrown around between all disciplines. I can speak from experience that many economists snear at any discipline that does not use mathematics extensively (most areas of biology, chemistry, and even some engineering!). Thus the only discipline that meets with near universal approval is physics (and math of course).
Personally, I think the whole thing is silly. I would hesitate to dismiss a field entirely given how little I understand of so many disciplines. Maybe this whole thing is moot anyway. Guys like Fair are basically mathematicians with extensive training in economics.
Ok enough sticking up for him. For kicks, take a look at this site.
http://fairmodel.econ.yale.edu/vote2004/index2.htm
You can check out updates and more importantly play with the model.
Try entering in your own values and see what it computes. This excercise made me a Doubting Thomas.
I entered in a -2% growth for real GDP per capita.
5% for inflation
and 0 for good news.
Prediction? Republicans win by a hair!
Give me a break. We just can't win!