Jim G on the Florida returns

Date: Wed, 03 Nov 2004 12:19:43 -0600

From: Jim G

Subject: Re: Meta-analysis website (fwd)

We were both wrong. I guess, as it turns out.

I hope you take back the Florida accusation. I've looked at the numbers, county by county. I saw this happening fairly early in the evening.

It was a fair election, and if the Democrats are to improve their chances next time, they have to accept that this is what the majority wanted - even if it is a fiercely divided country.

Here's why I don't suspect Florida mischief:

1) The turnout was uniformly high across the state, compared to 2000. The minimum was an increase of 6.3%, in Lee County (Fort Myers area), which was 59.06% for Bush in 2000, 59.76 last night. The maximum was an increase of 49.6% in Gulf County (rural Panhandle) which had only 8900 votes total, but went 59.7% Bush in 2000, 67.09% last night.

47 of the 68 counties had increases in turnout of between 20 and 40 percent from 2000, only two below 14 percent and two above 45 percent. There's no turnout smoking gun of either hidden or extra ballots. And nothing stood out during the evening when I was entering numbers.

2) Bush's vote percentage (calculated only by adding Bush and opponent votes, so Nader's not much of a factor here) was higher in 2004 in 60 of 68 counties. Four of the counties where Bush did worse were Republican, four were Democrat. Only one county (Gadsden - Tallahassee) saw Kerry do better than Gore did in 2000 (70.09% versus 67.12%).

Bush did better than a 5.5% improvement in 16 of the 68 counties, the largest of which was Sumter (south of Ocala) which saw 31,000 of the state's 7.3 million votes. Sumter went 55.72% for Bush in 2000, 63.08% in 2004. And saw unusually high turnout (+44%). But Sumter stands out in this manner, and the extra 4,000 Republican votes from the average increase you'd expect doesn't even begin to dent the 377,000-vote margin in Florida.

So, if there were fraud, it had to be subtle in adding uniformly to nearly every county. And then surely there would be some local officials screaming by now that totals weren't what they reported.

Also, moving to the Florida differing from the polls issue...

I tracked this carefully, compiled averages in much the same way you did. Florida was averaging +2% for Bush during the last week of the campaign. It finished +5%.

There were 16 states which had greater polling errors in Kerry's favor. In battlegrounds, here's what I had in terms of errors:

Arizona: K +4

Arkansas: K +6

Colorado: K +3

Florida: K +3

Iowa: K +1

Maine: 0

Michigan: 0

Minnesota: 0

Missouri: K +3

Nevada: 0

New Hampshire: K +2

New Jersey: B +2

New Mexico: B +1

North Carolina: K +5

Ohio: K +2

Oregon: K +1

Pennsylvania: 0

Tennessee: K +8

Virginia: K +2

West Virginia: K +5

Wisconsin: B +2

You can see that the polling was pretty good, but estimated Kerry about 2% better than he really did (3% in the closing moments of the race). Florida fits right in with that, almost exactly.

Note from Sam Wang: those numbers don't match entirely with my scatter plot (top of the validation page), which shows six states above the diagonal and nine below. Polls per state are included in Jim G's analysis. Also, in some cases our numbers disagree.

Perhaps this was simply an overestimate of the youth vote by the polling organizations, nothing more, nothing less.

I hope you reconsider your closing remarks. I've really enjoyed your site and I hate to see you finish with such a conclusion.