This shows how polls get it wrong--The Des Moines Register

[mistertrickster]

showed Kerry and Edwards in low 20's--they actually got high to mid 30's, way out of line for margin of error.

<http://desmoinesregister.com/news/stories/c4789004/23244233.html>

[XNASA]

1. This is misleading. It works different in a caucus.

For example, if in a particular caucus, Gephardt had less than 15%, then those voters would most likely switch, say to Kerry.

The way that people vote in the Iowa "widens the gap" between the candidates at the top and at the bottom. There's no way to take that into account in the polls leading up to the actual voting.

[Stuckinthebush]

2. I have to back up what XNASA is saying

The caucus is a strange event in politics. Don't judge polling accuracy by Iowa. To look at polling accuracy, grab a bunch of old polls from the 1996 presidential election. Plot these results on Excel with the margin of error noted. Then, take the actual results and plot those on the same graph, but as a different color. See how many of the polls are within or very near the margin of error.

Polling is an interesting science. If the sampling is done well, then you do have a very good indicator of public opinion.

[mistertrickster]

3. Okay, fine, but then the paper should have had a caveat saying

that their polling data doesn't have much reliability to caucus outcomes. Instead they ran it as page one with big happy pictures of Edwards and Kerry with red UP arrows and a less flattering picture of Dean with a big red DOWN arrow.

Poll or self-fufilling prophecy?

[Yo_Mama_Been_Loggin]

4. The thing I love about the results of the Iowa caucus is how stupid it ...

...made the punditocracy look. These are the same talking heads who predicted Clinton was through after his affair with Lewinsky. These guys should be locked in a room and given a severe ass kicking till their self important ego is finally deflated.