Monday, October 13, 2008

Scratch

For quite a while I've been covering the political horse race, but I think it's time to stop. One of the horses has pulled up lame. Unless some sort of catastrophe happens to the Obama campaign, I don't see how McCain can pull this out.

Some have noted that McCain has narrowed the gap in the Gallup Tracking Poll by four points over the last few days, and Zogby has Obama up by only 4%. I'm not impressed.

First, I believe these polls are accurate plus or minus 3%, and then only 95% of the time (assuming they're sampling correctly--and there's more reason to believe they're undercounting Democrats than the other way around, not to mention all the early voting probably helps the poll leader). So it's normal some are going to be a bit high or low. In the past few weeks, Zogby's been showing a closer race than the others, and, in fact, the 4% difference represents a larger gap than Zogby had not too long ago, so those numbers are trending toward the others. Meanwhile, Gallup's gap, which was 11% at the highest, has simply settled back down to "only" 7%.

The point is, Obama's lead seem to be somewhere in the 5% to 7% range. It rose to that a few weeks ago when the economic crisis started, and it's stayed there since. And it's unlikely the public is going to forget about the crisis before November 4.

If Obama were consistently 3% or less ahead, he'd still have a clear edge, but would seem to be within striking distance for McCain. But Obama is solidy ahead, and might even show better than the polls indicate, so it doesn't leave much room for a McCain victory scenario.

As far as changes in the poll numbers, my only suggestion is look at all the national polls, and if 75% or more of them show a greater than 2% change in the same direction over a 3-day period, then you've got a trend. The rest is noise.

3 Comments:

Blogger QueensGuy said...

Agreed. Also, the composite polls at fivethirtyeight.com and realclearpolitics.com are particularly useful for weeding out one pollster's errors and/or statistical outliers.

6:39 AM, October 13, 2008  
Blogger New England Guy said...

1980- Carter vs. Reagan was too close to call and Newsweek prepared a "deadlock" cover (which seems odd because Anderson was consistently polling under 10% in every state- they must have foreseen a 269-269 tie)

Yogi Berra Cliche.

1:43 PM, October 14, 2008  
Anonymous Anonymous said...

Those were the days when everyone voted on Election Day. Now, with a fifth to a third of votes cast before then, people won't have time to concentrate and fall away from one candidate.

3:28 PM, October 14, 2008  

Post a Comment

<< Home

web page hit counter