Saturday, October 16, 2004

Groundhog: Methodology

I received the following e-mail from George Axiotakis, who goes by "The Groundhog":

I am writing today, at the prodding of friends, to ask that my weekly update be included in your survey. At the risk of hubris, I do possess an unusual talent for predicting elections. Yeah, okay--this and $2 will get me on the subway--but my track record speaks for itself. (I was a history and anthro major who was and is fascinated by the electoral college and what it tells us about our evolving political culture). In 1992, I got four states wrong (I flip-flopped Ohio and Texas because of Perot). In 1996, it was three states (Arizona, Colorado and Georgia). In 2000, I only erred on Delaware and West Virginia (of course, I don't think I got Florida "wrong.")

I knew Gore would lose Tennessee, and I seem to be the only one who predicted that Gore would win New Mexico. That's better than John Zogby, Rasmussen or any of these clowns with computers.

How does one do it? Most pundits pay attention to the moving average of selected polls, and to a lesser extent, the results from the last election (that is why, for example, most experts do not think Bush has a real shot in New Jersey, or Kerry in Virginia, despite what any one poll might say). Professional pollsters will admit to one potential wild card: more younger people use cell phones as their primary phone, and polling firms do NOT call cell phones (but it should not be assumed that these same "young" people are reflexively liberal--WHERE are they?)

I employ two more factors, one quantitative, the other qualitative. The first is to gauge the potential effect of [Nader] who, with Republican help, has managed to land on over 40 state ballots. This could be crucial: for example, he is the only reason I can't now predict that Kerry will win New Hampshire or New Mexico. I believe that Zogby and Larry Sabato are right when they say that Ralphie will not equal his 2.7% of 2000, but it also depends on WHERE.

The second factor is more subjective, but related: to try to interpret the evolving political cultures of the states, and how these affect voting patterns. Of course, a disquisition on this is beyond the scope of this e-mail. But states do have political cultures; for example, if Latinos in the southwest voted at the same rate as blacks in Mississippi, the Democrats would sweep CO, NM, AZ and Nevada (but they don't). Suffice to say, there are many states that vote the way they do in presidential elections almost by rote; until 2000, a notable example was West Virginia. But the Republicans were able to win West Virginia in 2000 because they finally learned how to swing enough voters in a socially conservative state. (Conversely, until 1964 Vermont was the most Republican state--but thanks to migration, those days are gone.)

Now many writers have pointed out that the Democrats have done a better job than Republicans at registering new voters, especially in the swing states of OH, PA and FL, and the southwest (Republican strategists admit as much). Pollsters have no real way of assessing the effect yet--and nor does anyone else until election day. But the same pollsters have an odd tendency to undercount minority voting strength in urban states, and to believe the polls in southern states (many southern whites seem even more reluctant to share their real feelings with pollsters--often about black candidates. Yet virtually no pollster saw that Max Cleland would lose in Georgia in 2002--as my man Harold points out, there are those surveyed who would simply not admit that they weren't going to vote for the disabled guy). What this means is: if the final poll shows Kerry +1 in PA, bet on Kerry--he'll win by more than that; if another poll shows Kerry +1 in ARK, I'd be worried.

So what does all this portend? I am ready to predict Pennsylvania, MIchigan, Minnesota and Oregon will go for Kerry (total 55 EV); the polls show small but consistent leads. For the same reason, I am prepared to concede Virginia, Arkansas, Colorado and Arizona to Bush (38 EV, pending the outcome of Colorado's amendment 36). I will not attempt an "if the election were held today" prediction for EVERY state, because the election is not today. BUT as of today 10/12/04:

(17 states + DC) (24 states)

That leaves nine tossup states with 94 EV: Florida(27), Ohio(20), Missouri(11), Wisconsin (10), Iowa(7), Nevada(5), New Mexico(5), West Virginia(5) and New Hampshire(4). (Okay--For those who care, as of today OH, WI, NH and NM lean slightly Kerry; that would make his total 238+39=277. BUT THE ELECTION IS NOT TODAY.)

You can view my map (and other prediction maps) at Dave Leip's Atlas of US Elections,

Go to 2004
Search for prediction by username
Mine is "ground_x"

Ed Fitzgerald | 10/16/2004 11:23:00 PM | write me | | | HOME | MAIN SITE

Ed Fitzgerald

Clowns to the left of me,
Jokers to the right,
Here I am...
back to main site
original content
copyright 2004 - 2008
Ed Fitzgerald


take all you want
but credit all you take