Polls in close elections are hardly ever accurate.

[Sean Reynolds]

Why? Because in a close election it's easier to oversample candidate A's supporters over that of candidate B's. In a landslide election polls tend to be a bit more accurate because a wide margin of people support the winning candidate.

Example: Room filled with 10 people. You're supposed to poll only 6 of those 10. However 5 of those 10 are supporters of candidate A and the other 5 are supporters of candidate B. You poll 6 but it breaks down this way: 4 for candidate A, 2 for candidate B. Candidate A wins the polling, albeit by a close margin. However is that reflective of the whole room? No. So unless you poll all 10 people (or the entire country in this case), you can't get a full understanding of where people stand in a close race.

You can use this example for a landslide as well. 10 people in a room, but this time 8 people support candidate A, while 2 people support candidate B. You poll 6 of those 10 and you find that 5 support candidate A while only 1 supports candidate B. Candidate A wins, and it's reflective of the room as well, correct?

My point? Polling in close elections can be off 3-4 points because you can either sample too many supporters of one candidate, or too few supporters of one candidate. When you do that in a close race it doesn't realistically give you the makeup of the race. Take 2000 for example. Most of the polls had Bush winning by a margin of 5-10% points. Only a few polls had Gore winning the popular vote. In the end Bush LOST the popular vote and Gore won it. Granted the margin was small, but it was still quite different than Bush winning by 5 or 10 points.

Another example locally. The Salt Lake Tribune had a poll last November for our mayoral election. In THEIR poll they had Rocky Anderson LOSING to the guy running against him. He was losing by like one point. The Deseret News released a poll having Anderson up by 2 or so. The election came and Anderson won by 7 or so points...a strong showing for a 'supposed' tight race.

So when you see a poll, don't trust it. Whether it shows Kerry winning or Bush winning isn't the issue. The issue is that this is a tight race, and tight races don't favor incumbents......

[Mz Pip]

1. Olbermann did a segment on polling

tonight. Pollsters are only getting about 20% of those polled answering all the questions or cooperating. I wonder what that does to the numbers?

MzPip
:dem:

[Xipe Totec]

2. The ones that aren't close, aren't accurate either