2000 Nader Vote by County in MN versus swing to Kerry from Gore

[Torie]

Below is the scatter diagram, with the X axis the 2000 Nader percentage by county in Minnesota, and the Y axis is the percentage swing towards Kerry as compared to Gore's percentage. As one can see there is a correlation and a trend line in favor of the hypothesis that the higher the Nader percentage, the higher the swing, but it is very loose, and the R squared factor is only .07, which is tiny. So a LOT more was at play in the swing than the size of the Nader vote.

I am doing this because some of the usual suspects were marginalizing me by suggesting that I was all wet in the notion that Oregon might be close this year, because the only reason it was close in 2000 was Nader. Having gotten beyond themselves, it was time to take off the gloves.

[Sam Spade]

Minnesota may a be a different beast than Oregon concerning Nader, just FYI.   

The Dem GOTV effort in 2004 simply swamped the GOP effort in MN - otherwise MN would have been much closer.  In fact, most Bush people thought the state was a complete wash going into election eve - they ended up behind by 3.5%

[Torie]

Minnesota is better because it has more counties, and thus more data points. And the size of the trend to Kerry should not affect the degree of the scatter (the R squared factor); rather  just where the intercept is, and the  slope.

[Verily]

Yes, but Oregon is different. Especially so vs Minnesota, which has its own quirky brand of independent supporters not found in many other states. Oregon, of course, has a lot of hippies and hipsters who supported Nader but always vote Democrat. Minnesota has a rebellious streak, but the rebellious voters are not uniformly Democratic when no viable third party is available (while they generally are in Oregon).

[Torie]

Well to appease the clammering throng of tendentious enfant terribles, I will "do" Oregon in due course.

[Beet]

This only works if you assume a uniform swing: that the only variable determining the swing between 2000 and 2004 is Nader. Of course it is not going to explain all the variance. But disproving that hypothesis does not disprove the hypothesis that "the only reason [Oregon] was close in 2000 was Nader."

To take some extreme simplifications, suppose that in counties A and B, the vote in 2000 was 45% Bush, 45% Gore, and 10% Nader, while in C and D, the vote in 2000 was 60% Bush, 40% Gore, 0% Nader. In 2004, counties A and B became 55% Kerry, 45% Bush (a swing of 10), while C and D became 55% Bush, 45% Kerry (a swing of 10). In a regression model, these numbers would not create any sort of correlation whatsoever; but an interview of the Nader supporters in counties A and B could very well have produced a result of 80% of them saying Gore was their second choice and they would have voted for him.

In short, proving that Nader was not the only variable leaves open the possibility of arguing that a combination of pro-McCain forces stronger than those for Bush in 2000 or Kerry in 2004 could make the state close, but does not allow you to argue that a pro-McCain wind similar to that which carried Bush in 2000 would make the state close.

[Torie]

That is all splendid Beet if one assumes that both 1) Nader voters are Gore/Kerry voters, and 2) the fewer Nadar voters in a county, the more Bush 2000 voters went to Kerry. If true, that would all be extremely curious. But yes, one way to neutralize my data is to assume that there was some other factor behind the curtain that offsets the swing of Nader voters to Kerry, and thus on the surface makes it appear to disappear, but it's really there! 

[Beet]

That is all splendid Beet if one assumes that both 1) Nader voters are Gore/Kerry voters, and 2) the fewer Nadar voters in a county, the more Bush 2000 voters went to Kerry. If true, that would all be extremely curious. But yes, one way to neutralize my data is to assume that there was some other factor behind the curtain that offsets the swing of Nader voters to Kerry, and thus on the surface makes it appear to disappear, but it's really there! 

I'm not trying to "neutralize" your data, but what you're essentially doing is trying to infer on individual behavior (how someone votes) by measuring variation between geographical units. What is missing here is the possibility of variance within the units.

It is fairly obvious that

(1) the 2000 and 2004 elections produced non-identical results for reasons different from Nader's smaller presence in 2004. For one thing, the Democrats had a different candidate; for another, the issues were completely different. Not many people would argue that these factors did not influence votes in all 50 states.

(2) Nader's vote is non-identically distributed. In some counties, he gets over 5% of the vote; in other counties, just 2%. It is obvious that in the counties where he got a smaller percentage of the overall share, his impact on the final outcome in that county is going to be smaller. A hypothetical county where he got a 0% share of the vote, he could not possibly explain any of the variance.

In order to make definitive statistical claims about Nader's impact (or non-impact) on the Oregon results, you would have to conduct interviews of Nader voters themselves and ascertain their hypothetical intentions. That is all I am saying.

[Torie]

Yes, that is why a lot of data points are good. Either there is a pattern or there isn't, and then one can speculate why. I already said there were polls that had 40% of Nader voters going to Gore, and 20% to Bush, but that was greeted with skepticism. There is this desire to unduly demonize Nader for the ascension of Bush.

[Torie]

The Nader thing does appear to be more robust in Oregon. I hate when that happens! More at 11, or something.

[Sbane]

The Nader thing does appear to be more robust in Oregon. I hate when that happens! More at 11, or something.

Yup OR hippies are much different from Minnesota independents. You can notice the same thing in Humboldt and Mendocino counties as well. Very heavy swing towards Kerry and not all of that can be explained by anti-war sentiments.