Every poll? Geez, what have you been watching? The polls keep swinging wildly around; in the past two weeks, we've had a poll show Bush with a ~10 point lead, then Kerry with a ~2 point lead, then Bush with a ~10 point lead, then Kerry with a ~2 point lead again. Checking the news today, they're now saying
Bush has a 3 point lead. Here in Iowa, the same sort of thing happened on the stateside polls.
While it's normally not this bad, phone polls have some serious problems (and they're getting worse, since cell phones aren't called and more and more people are going without land lines). The biggest problem is the response rate. It's a dirty little secret in the polling world that at least a third of the people that you call typically refuse to repond. This is a huge statistical problem, because it skews your sample. For example, if single mothers tend to be busier than the average person, they might be less likely to take phone poll calls, and you get your sample weighted against single mothers. If young 20-something men want their voices to be heard more, they may be more likely to take phone polls, and their views get disproportionally represented. Etc.
Foreign phone polls, especially in third world countries, are typically little more than PR (I remember the Venezuela polls saying that 80% of the country wanted to vote out Chavez (who won easily in the recall); articles reporting the result typically didn't mention that only 20-30% of the country's people - typically the wealthy, who hate Chavez - have phones).
Foreign door to door polls are usually better, but still not great. For example, look at the CPA's polls in Iraq. I remember one CPA poll, conducted door to door that was supposed to represent the whole country, showed notably more sunnis than shiites. That's a big demographic problem there
Polls have their uses. But until the issues of accessability and the refusal rate can be addressed, they're not that great.
"Illuminant light,
illuminate me."