Why The Polls Are Wrong
Math time, boys and girls!
Here's what the pollsters and the news media won't tell you. Here's why the polls are wrong.
Get out your statistics textbooks. Let's turn to the section on "total universe." The larger the market, the more accurately you can forecast its behavior, because the easier it is to get a representative sample.
In a national election, the total universe is over 100 million, so it's relatively easy to get a representative sample.
In a New Hampshire primary, the total universe may be 100-150,000. We don't know. Not only that, but because the total potential universe is small, it's very hard (nearly impossible) to draw a representative sample from it. All the "tricks" pollsters use to even out differences among samples, which work well with a big universe, skew the results further with a small universe.
In a national election, where 100 million vote, you need to change 1 million minds to get a 1% movement in the polls. In a New Hampshire primary, you need a change of 1,000 minds. You can get the same impact by changing the contents of the sample, by changing who actually turns out.
So there is no way for the polls to be right in New Hampshire, and frankly, they never are. They are nearly always wrong. Last time, in the Republican primary, polls taken the day before the election showed Bush up 4% on John McCain. He lost by 17%.
What does this mean? First, momentum is important, and right now it appears Dean has it. Second, Get Out The Vote (GOTV) activities mean everything. You've got to get out all the voters you can. Then you hope for the best.
Fingers crossed.
Here's what the pollsters and the news media won't tell you. Here's why the polls are wrong.
Get out your statistics textbooks. Let's turn to the section on "total universe." The larger the market, the more accurately you can forecast its behavior, because the easier it is to get a representative sample.
In a national election, the total universe is over 100 million, so it's relatively easy to get a representative sample.
In a New Hampshire primary, the total universe may be 100-150,000. We don't know. Not only that, but because the total potential universe is small, it's very hard (nearly impossible) to draw a representative sample from it. All the "tricks" pollsters use to even out differences among samples, which work well with a big universe, skew the results further with a small universe.
In a national election, where 100 million vote, you need to change 1 million minds to get a 1% movement in the polls. In a New Hampshire primary, you need a change of 1,000 minds. You can get the same impact by changing the contents of the sample, by changing who actually turns out.
So there is no way for the polls to be right in New Hampshire, and frankly, they never are. They are nearly always wrong. Last time, in the Republican primary, polls taken the day before the election showed Bush up 4% on John McCain. He lost by 17%.
What does this mean? First, momentum is important, and right now it appears Dean has it. Second, Get Out The Vote (GOTV) activities mean everything. You've got to get out all the voters you can. Then you hope for the best.
Fingers crossed.
Comments