Posted on 06/16/2004 8:17:00 AM PDT by Darth Reagan
THE OSCILLATING RACE
Bush 47.17% - 256 | Kerry 48.66% - 282
June 15, 2004
Last week, Kerry lead the composite poll by 1.54% and this week, he leads by 1.49%, but the electoral vote picture has switched from last week's Bush lead of 281-257 to a Kerry lead of 282-256. While the state by state picture continues to oscillate between the candidates, there's been a clear trend in Kerry's favor in the national composite results, but that trend has slowed and for the last month, there hasn't been much change.
This week's composite poll includes the suspicious LA Times poll, showing a Kerry lead of 6 points. I discussed why it is suspicious here. Without that poll's results, the composite poll would show a dead even race.
This week's analysis includes 9 new state polls for Illinois (Kerry +13), Kentucky (Bush +13), Michigan (Kerry +2), Missouri (Bush +11), New Hampshire (Tied), Ohio (Kerry +3), Pennsylvania (Kerry +1), West Virginia (Kerry +6) and Wisconsin (Bush +2).
Bush State Gains
None
Kerry State Gains
Ohio
West Virginia
Toss Ups (slight lead of less than 2%)
Florida (Bush)
New Hampshire (Kerry)*NEW*
New Mexico (Kerry)
Ohio (Kerry)
Oregon (Bush)
Pennsylvania (Kerry)
West Virginia (Kerry) *NEW*
The States
Margin of Lead
(+Bush | -Kerry)
|
Alaska | 27.52% Arizona | 4.73% Arkansas | 3.58% California | -8.34% Colorado | 7.27% Connecticut | -11.43% Delaware | -15.29% District of Columbia | -79.78% Florida | 0.14% Georgia | 14.08% Hawaii | -21.61% Idaho | 38.25% Illinois | -13.07% Indiana | 19.85% Iowa | -5.32% Kansas | 18.64% Kentucky | 10.53% Louisiana | 12.01% Maine | -14.40% Maryland | -13.92% Massachusetts | -24.99% Michigan | -2.37% Minnesota | -4.34% Mississippi | 15.36% Missouri | 5.83% Montana | 20.27% Nebraska | 26.73% Nevada | 4.88% New Hampshire | -1.18% New Jersey | -7.60% New Mexico | -1.54% New York | -21.23% North Carolina | 5.78% North Dakota | 26.01% Ohio | -0.20% Oklahoma | 21.75% Oregon | 0.23% Pennsylvania | -1.36% Rhode Island | -32.37% South Carolina | 10.42% South Dakota | 16.73% Tennessee | 6.44% Texas | 17.23% Utah | 44.05% Vermont | -13.40% Virginia | 2.40% Washington | -5.34% West Virginia | -1.05% Wisconsin | -1.61%
|
There are too many significant figures displayed relative to the input data. Displaying (in this case) four significant figures implies that the data is good to 1 part in 10000. Most of this data is good to about 1 part in 30. This is a common advertising ploy; the purpose is to give the data more "gravitas" by implying a high precision. There's a PDF file discussing some of the issues here .
I think that there is an error in your formula. You are dividing by the Aggrate Poll but I think you only meant to divide by the weights (I may have read it wrong). The formula is prima facia wrong as it does not reflect the correct method of combining statistical data. This isn't too importang because your weighting of polls has a higher impact.
Considering that 2000 was a good Democrat year, and with the 72-hour plan, 2004 should be a bit different, and better for Republicans.
That's funny. The reason that I constructed formulas is to get my bias out of it. Otherwise, I'd show what I really think, which is a Bush win by no less than 30 electoral votes. In fact, there may be a basic bias in the formula and my unwillingness to exclude the LA Times poll.
If the LA Times poll is really aberrant, I think it is, then this analysis will show a more dramatic shift toward Bush next week. Although it may take a while for the states to swing back since I weight the more recent polls more heavily.
Yes I do. Today, they are just wondering what's wrong with my Wisconsin numbers.
Check below:
Wednesday June 16, 2004--The latest Rasmussen Reports Presidential Tracking Poll shows President George W. Bush with 46% of the vote and Senator John F. Kerry with 45%. The Tracking Poll is updated daily by noon Eastern.
Thirty-nine percent (39%) of the Tracking Poll sample are Democrats while 35% are Republican. These numbers are consistent with the partisan turnout in the last Presidential Election.
http://www.rasmussenreports.com/Presidential_Tracking_Poll.htm
I'll look again at the formula that I summarized on the methodology page. My description of weighting is correct, but I may not have summarized the formula properly.
What's the correct way of combining statistical data? I admit to being an amateur (Dammit, Jim, I'm a lawyer, not a statistician). I wanted to do more than a straight average because I believe Likley Voter polls should be weighted more heavily than Registered Voter polls, or all adults.
I'm not sure I understand what you are analyzing in your post #16.
I'll post more later when I have time. When combining means, you need to weight inversely according to the variance of each mean. You may be ok with what you're doing. Roughly the variance of a poll is proportional the number sampled so you're doing that right. Modifying for likely-registered is ok too (as long as it's done consistently and noted in the results.) I think there's a minor modification I'll suggest later.
Also, if states are close (within a few percentage points) you should just call them a tossup.
BTW FWIW I agree with Dr. Stochastic's comments regarding the precision of the data as it's presented. The number of significant digits displayed does imply the data presented is accurate to that level.
The L.A. Times polls will not have much of an impact on your results because they don't do many of them, but the ARG and Rasmussen polls will. Your site can be seen as a sort of worst case scenario for Bush (or best case scenario for Kerry).
If I were smarter, I might be able to put together a site which had the polls normalized. That's a worthwhile project, which maybe I'll try if I can figure out how to log on to my computer. I also have to figure out make the pretty pictures with the EV's.
If you wanted to come up with a normalized map, that would be pretty cool. (One could choose which poll to use as the base line, and the map would display according to that).
How would you do that? Again, not sure I am following. How would you go about selecting a poll as a baseline?
Sorry, I was in a hurry for a meeting, and wasn't very clear. The numbers represent how much you have to add for Kerry to normalize the data to SUSA. So for Rasmussen, if you add 3.9 you get a result which would be comporable to what SUSA would be expected to show.
I hope that's clear. If it's not, I can explain further.
I'm using SUSA as the base line poll because it was one of the most accurate state polls last time (second behind Mason-Dixon, if I'm not mistaken) and they have quite a few polls out. Also in independent analysis I've done on the election, so far in 2004, they have been the most accurate (this is on the average of course -- any given state can have an odd result).
Oh, now I get it. That makes sense. Of course, selecting SUSA is a shot in the dark -- based on historical data which may or may not be indicative of future results.
For national polls, the same method would require picking Zogby as the baseline - based on 2000 results. Which may have been more about luck in guessing the partisan turnout.
"...GWB is going to win big..."
Yup. Just wait until the debates. It's gonna be great to watch JFK squirm and stutter. hehhehheh
Eh, who cares? We're obviously dealing with averages of numbers with their own margins of error, so I don't see anyone being led astray. Besides, the odds of the actual result being a whole number (52 to 48) are the same as being a particular decimal (52.12 to 47.88). Just because the tenths and hundredths digits in 52.00 aren't shown doesn't mean they aren't there. It's just psychological.
In a previous post you suggested SUSA was a shot in the dark. Here's how I settled on SUSA.
1) The 3 most accurate state polls last time around were Mason-Dixon, SUSA and Resarch 2000. Only SUSA has done a significant number of polls this year. Choosing one of the other two would lead to too much possible random fluxuation. SUSA has done enough polls that an odd result (such as the CA poll showing Bush ahead by 1) doesn't affect their overall average very much.
2) Aside from the 2000 results, my understanding is that SUSA has a good track record
3) Before I researched into SUSA I had been doing my own tracking of the state polls, based on an independent methodology, and the SUSA results were most in line with what I was showing. So even without the past results, if I were to choose a poll based on what I was showing, it would have been SUSA.
Regarding the normalization, my idea is to allow the user to choose whatever poll he wants as the baseline. So if a user wants to use Rasmussen as the baseline, he could. The idea is to normalize the polls so they're consistent with one another. Without the normalization a state-tracking site is at the mercy of whatever poll happens to be most recent.
Bush cannot depend on the debates, that would be dangerous IMO.
Here's some normalized results. First normalized for SUSA:
1) FL +7
2) IA -4
3) MI -4
4) MN -1
5) MO +5
6) NV +12
7) NH +4
8) NM +5
9) OH +6
10) OR +5
11) PA -1
12) WA -5
13) WV +9
14) WI 0
15) CA -4
16) CO +9
17) NC +8
18) VA +6
Now normalized for Rasmussen:
1) FL +3
2) IA -8
3) MI -8
4) MN -5
5) MO +1
6) NV +8
7) NH 0
8) NM +1
9) OH +2
10) OR +1
11) PA -5
12) WA -9
13) WV +5
14) WI -4
15) CA -8
16) CO +5
17) NC +4
18) VA +2
You can see how the second group of numbers looks more what like what we're used to seeing. This is because Rasmussen does so many polls.
Since the polls like ARG, Rasmussen and Zogby do more polls and are favorable towards Kerry, a picture is presented of Kerry being ahead. There are no polls which are favorable towards Bush in the way the above polls are favorable towards Kerry. The SUSA polls present a neutral picture.
This is supposing that the Dem favoreable assumptions used by Rasmussen and others (e.g. high Dem turnout) are incorrect. If they are correct, then it's possible SUSA will be wrong this time around and the others will be right.
The SUSA state polls correspond roughly to the national polls of IBD, Battleground and Fox while the Rasmussen state polls correspond roughly to the Zogby and ARG national polls.
The first list (normalized for SUSA) would give Bush a lead of 290 to 207 with WI, MN and PA as toss-ups (I'm calling anything within 2 points a toss-up).
The second list (normalized for Rasmussen) would give a 248 to 248 tie with OR, OH, MO and NH toss-ups.
What we see on Dales and your site is something in between these two lists, but more towards the Rasmussen list because more Rasmussen polls are done.
If the polls favoring Bush and Kerry were equally inaccurate and equal in number, you'd get a decent picture of things, but the polls favoring Kerry are both greater in number and in quantity, so we get a skewed picture.
The media is just trying to make it a race and sell ads. They also hate Bush. But Bush will win hands-down.
The LA Times poll is included in the composite. Worthless.
I know that this question is going to make me look like an idiot, but what do you mean by normalize? Is it a comparison of state polling to certain national polls to determine if they are tracking similarly? It seems that you are saying that if you take the 2000 results, add the national poll, then the particular state polling should look a certain way. Is that it?
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.