Why do people assume polls are biased

[Foggyhill]

1. stats are science but the stat

Part is meaningless if the sampling part is not representive of the underlying pop your trying to poll

Yes, the stat part is meaningless if the polling is badly done

The polling methodology part has nothing to with science at all. There are best practices to attempt to get a random sample but none can guarantee it will.

You poll land lines and you only get a good idea of people responding to polls on land line will vote and the stat analysis for those people would be sound. The problem comes when saying those results apply to the general population... they do not. That's a big lie and is misrepresentation of the results.

This apply to any other methodological bias, like calling cells only, calling more in the south, more in the suburbs, more rural, calling time, order of questions, etc.

As for most polls being done in good faith, I don't think so. Any national poll of less than 1500 cannot truly be called that. You read their methodology and it is obvious they're not even following best practices (because getting a random sample is not yet science)

All the various methods introduce potential bias, likely voter models, difficulty and expense in insuring a random sample means even in the best of case, the margin of error is larger than mere stats for the sample size would dictate. This unknown error is unknown in size though it itself could be estimated (but that's a guess mostly, not scirnce)