Pollster.com

Articles and Analysis

 

Actual Vote Margin by Pre-Election Trend Estimate


votebytrend.png
The plot shows the relationship between our pre-election trend estimate and the vote margin. These are not quite complete data but as of 8:20 AM Wednesday. But I don't  expect large changes. DC is omitted because there was virtually no polling there.

Which states are which you ask? Behold:
votebytrendlabels.png
How about quantifying the errors? Here is the distribution with descriptive statistics in the plot.
trenderrordistribution.png
So on average we missed the margin by less than half  percentage point. Most states we got within five points. And a few we missed b more than five points on the margin (or two and a half points or more on each candidate.)

We'll be doing lots more looking at poll performance and how our estimators succeeded and where they failed. But probably not before a nice restorative nap.

 

Comments
Adam:

I made a similar scatterplot using RCP's numbers and found the same thing: the predictions tended to underestimate both McCain's support in red states and Obama's support in blue states; in the swing states, it was awfully close.

____________________

Adam:

And it's worth noting, by the way, that on the scatterplot, you can consider anything in the upper-left or lower-right quadrant to be a true red-to-blue or blue-to-red "surprise." NC's right on the line, so Indiana's really the only one.

____________________

cinnamonape:

I'm wondering if the high variance in the very blue and very red states was likely a consequence of the relatively few surveys there, thus allowing methodological errors of a few polls to underestimate the results. Of course, I suppose that they could have overestimated the results as well, so it's interesting that they tended to make it more of a "horserace" than it really was.

____________________

RWCOLE:

There's one oddity in the election returns so far. Stevens was 20 down in the recent poll after his conviction and now may have won re-election. Anyone have a reason for this?

____________________

RossPhx:

I agree with cinnamonape -- the fewer polls (because the state is not a battleground), the less reliable the consensus.

____________________

Adam:

@RWCOLE:

Maybe the Republicans suddenly realized that a vote for Stevens was most likely a vote for a special election (assuming the Senate chucks Stevens out) that might be won by a different R.

____________________

geraldatwork:

I am also curious about the Stevens victory in Alaska. He was down a few points in the polls and if I remember correctly the last poll before the election showed him down about 8-10 points. Did Palin go up to Alaska to help fix the machines. I would like to see exit polls seeing who they voted for and compare it to the result. I smell something here. If he was kicked out of the Senate she could appoint herself Senator a much better position going forward to the 2012 election.

____________________

zealotus:

@gerald

I truly doubt there was a conspiracy... I think Alaska is just a particularly idiosyncratic state that is a little more difficult to poll accurately. Particularly when the average Alaskan may feel like the lower 48 are treating them as the butt of some sort of Palin/Stevens/Young joke, they may have just decided to stand up for their fellow Alaskans at the expense of voting for a convicted felon.

____________________

cinnamonape:

One more point - about Georgia. A substantial number of early voter results from Fulton and Gwinnett Counties are still uncounted. That could be as many as 275,000!

http://www.11alive.com/news/local/story.aspx?storyid=123175&provider=top

____________________

Lukas:

zealotus says: "Particularly when the average Alaskan may feel like the lower 48 are treating them as the butt of some sort of Palin/Stevens/Young joke, they may have just decided to stand up for their fellow Alaskans at the expense of voting for a convicted felon."

Hmm, perhaps. But if that's the case it is just going to lead to Alaskans being the butt of even more jokes.

____________________



Post a comment




Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.

MAP - US, AL, AK, AZ, AR, CA, CO, CT, DE, FL, GA, HI, ID, IL, IN, IA, KS, KY, LA, ME, MD, MA, MI, MN, MS, MO, MT, NE, NV, NH, NJ, NM, NY, NC, ND, OH, OK, OR, PA, RI, SC, SD, TN, TX, UT, VT, VA, WA, WV, WI, WY, PR