Recently, Steven Den Beste took a look at the overall poll trends in this race and spotted an overall trendline running through the data in this race. Den Beste is an engineer by trade, so he knows a thing or two about spotting mathematical trends.
My analysis of this election is remarkably similar to his:
What we see is a trendline in which George W. Bush consistantly leads John Kerry by a few percentage points each time. The Real Clear Politics poll average is a good measure as it smooths out some of the sampling bias in the polls and deals only with aggregate figures. Despite all the jitters in the data, the trendline remains consistant even with the polls that have come out subsequently to Den Beste’s analysis.
Will this trend hold? I believe it indeed will, and based on the polling that’s come out, I think this overall trendline is about right. However, I’m still not quite confident enough to make a firm prediction quite yet. If the recent Mason-Dixon polls (noting that Mason-Dixon nailed it in 2000) are correct, George W. Bush will have a convincing win so long as he hangs on to Ohio. The Buckeye State will be the key battleground in this race, and that’s the state I’ll be watching the most closely through Election Night.
More as I continue to crunch the numbers and see if I can get my political scientist’s crystal ball of political psychology, polling, and r-squared values to get a real prediction out as this race begins to solidify.
The polls that appear to have a genuine methodology, be it NBC, Rasmussen, Zogby, TIPP or the new Marist poll show the race dead even or an insignificant +1 for Bush. Most of the polls showing Bush leading tend to oversample Republicans, particularly Gallup. The only poll that concerned me this week is the Fox News poll. As messed up as the network is, their polls have traditionally been reasonable. However, their internals seemed questionable, so I’m cautiously dismissing it is a bad sample. Mason-Dixon seems to skew right, particularly this election sample. They very well might be right, but I’m yet to be convinced.
The bad news for Bush is that most polls indicate Kerry with a comfortable lead in the battleground states. The new Marist poll shows the race tied nationally, but with Kerry leading by eight points in the battleground states. Meanwhile, domestic issues are still cited as the main concern among battleground state voters, which benefits Kerry.
Your predictions of the race thus far have been anemic. Just a month ago, you ridiculed me for insisting Ohio was still a battleground state, despite Bush’s uptick in the polls. You were and continue to be convinced that the poll of the hour is indicative of how voters will feel on November 2, as evidenced by your repeated embarrassing aboutfaces on Ohio. I have always been less certain of this election’s outcome and am not ruling any scenario out. As it stands now though, I’d give Kerry 55-45 odds.
I’ll admit Ohio was one that trended away from Bush against the previous trendlines. States can and do swing towards a candidate then swing back.
And as I’ve said before, the poll of the hour is only a signal of larger trends, which is why I’m using the RCP averaging rather than polling data.
And BTW, using the raw polling data yields similar results. Using the raw polling data indicates a closer race, but still shows Bush slightly increasing in support and Kerry decreasing slightly over time. However, the r-squared values for Bush were only .22 and the Kerry r-squared values did not show any kind of significant correlation (I believe is was something like .06). This is why I didn’t include them here and why using raw polling data is too statistically incoherent for analysis.