Time to start a fight (hee-hee):
The Newsweek and CNN/USA Today/Gallup polls were conducted over this last weekend, which traditionally over-values Democrats. In comparison, weekday polls often give more support to Republicans, though there is no evidence that the Republican support is over-valued.
The CNN/USA Today/Gallup poll is not the same poll as the traditional Gallup poll. Over the course of this year, Gallup-only polls have favored President Bush more highly than the CNN/USA Today/Gallup polls, suggesting slightly different methodologies.
The Newsweek poll has a very strange methodology, which cannot be reconciled with established and accepted procedures. On Thursday night, Newsweek polled registered voters immediately after the debate, and in doing so, polled only the Mountain and Pacific Time Zones, and given the short time available, is likely to have concentrated on urban areas in the West, which is extremely prejudicial in favor of Kerry. Although Newsweek continued polling on Friday and Saturday, there is nothing to indicate that Newsweek balanced its geographical bias.
Newsweek changed its political party weighting to produce its desired results. While I am open to the concept that a poll will weight its respondents as it sees fit, to change the ratio of Republicans to Democrats to Independents from one poll to another is not valid as a practice.
For all the weird behavior, however, Newsweek produces details, which allows us to go back and see what was really going on. Looking at the last three polls, the following trends emerge:
#1. GOP support for Bush was at 94% on September 3rd, down to 93% on September 10th, and 89% on October 4th. The West Coast tilt accounts for a lot of that. The Democrats' support for Kerry is interesting, moving from 82% to 87%, then down to 86%. Kerry did not move his base. The real change is Independents, who went from 45-40 Bush, to 45-39 Kerry and now 42-37 Kerry. All-in-all, not a major change, which is not good for Kerry, though he actually reclaimed the Independents back from Bush about a week BEFORE the debate. For here, it's worth noting the weight of the polls:
- The September 3rd poll showed a headline of 54-43 Bush, followed by a September 10th headline of 50-45 Bush, and now 47-45 Kerry. If you check the unweighted numbers, however, and apply party support to them, they play out this way for unweighted results: September 3rd, 54-39 Bush; September 10th, 51-41 Bush; October 4th, 46-46 tie. Since the polling numbers seem to be suspect, that's not as good for Kerry as it sounds, but it also shows a different road to the moment.
- Next, assume for the moment that the actual turnout will be dead even; 33.3% for Republicans, Democrats and Independents. If that is the case, then the September 3rd position was 51-42 Bush, the September 10th position was 46-45 Bush, and is now 46-45 Bush if the sample polling data is accepted.
It isn't definite that the vote will work out this way exactly, but in years past, the turnout by party has bounced a bit. Double-digit ranges exist between election years and this one may well come down to who gets their butt off the couch and votes.
This one's courtesy of the folks over at PoliPundit. And it gave me a headache thinking about it :-)
Tony
The Newsweek and CNN/USA Today/Gallup polls were conducted over this last weekend, which traditionally over-values Democrats. In comparison, weekday polls often give more support to Republicans, though there is no evidence that the Republican support is over-valued.
The CNN/USA Today/Gallup poll is not the same poll as the traditional Gallup poll. Over the course of this year, Gallup-only polls have favored President Bush more highly than the CNN/USA Today/Gallup polls, suggesting slightly different methodologies.
The Newsweek poll has a very strange methodology, which cannot be reconciled with established and accepted procedures. On Thursday night, Newsweek polled registered voters immediately after the debate, and in doing so, polled only the Mountain and Pacific Time Zones, and given the short time available, is likely to have concentrated on urban areas in the West, which is extremely prejudicial in favor of Kerry. Although Newsweek continued polling on Friday and Saturday, there is nothing to indicate that Newsweek balanced its geographical bias.
Newsweek changed its political party weighting to produce its desired results. While I am open to the concept that a poll will weight its respondents as it sees fit, to change the ratio of Republicans to Democrats to Independents from one poll to another is not valid as a practice.
For all the weird behavior, however, Newsweek produces details, which allows us to go back and see what was really going on. Looking at the last three polls, the following trends emerge:
#1. GOP support for Bush was at 94% on September 3rd, down to 93% on September 10th, and 89% on October 4th. The West Coast tilt accounts for a lot of that. The Democrats' support for Kerry is interesting, moving from 82% to 87%, then down to 86%. Kerry did not move his base. The real change is Independents, who went from 45-40 Bush, to 45-39 Kerry and now 42-37 Kerry. All-in-all, not a major change, which is not good for Kerry, though he actually reclaimed the Independents back from Bush about a week BEFORE the debate. For here, it's worth noting the weight of the polls:
- The September 3rd poll showed a headline of 54-43 Bush, followed by a September 10th headline of 50-45 Bush, and now 47-45 Kerry. If you check the unweighted numbers, however, and apply party support to them, they play out this way for unweighted results: September 3rd, 54-39 Bush; September 10th, 51-41 Bush; October 4th, 46-46 tie. Since the polling numbers seem to be suspect, that's not as good for Kerry as it sounds, but it also shows a different road to the moment.
- Next, assume for the moment that the actual turnout will be dead even; 33.3% for Republicans, Democrats and Independents. If that is the case, then the September 3rd position was 51-42 Bush, the September 10th position was 46-45 Bush, and is now 46-45 Bush if the sample polling data is accepted.
It isn't definite that the vote will work out this way exactly, but in years past, the turnout by party has bounced a bit. Double-digit ranges exist between election years and this one may well come down to who gets their butt off the couch and votes.
This one's courtesy of the folks over at PoliPundit. And it gave me a headache thinking about it :-)
Tony