1. Polls are all biased. An unbiased poll is something of a work of art and they're difficult and expensive to prepare. Professional researchers allow for this in the analysis section, which is also expensive.
2. Polls are merchandise. They're something prepared for money sold to somebody who pays for them. If you're getting them for "free", meaning somebody else is paying for them, then that somebody is trying to sell you something.
3. Polls can be descriptive or normative, but not both. The cost for using them to influence behavior is that the results no longer work as objective descriptions. Political staffs tend to forget this and take their push polls' results seriously. It results in nasty surprises.
4. Polls are transient. They are a snapshot of answers given at a specific time and the longer it takes to gather and analyze the results, the bigger the error bars.
5. Online polls are worthless.
6. Polls about sexual behavior are worthless. Polls of teenagers concerning sexual behavior are hilariously worthless.
7. Polls whose answers may get the respondents in trouble with the law are worthless.
8. There is no such thing as anonymity.
9. Never confuse polls and data analytics. (Hillary Clinton's presidential campaign made that mistake). Past behavior is only an indirect predictor of future performance.
10. The best, most up-to-date, most accurate (and most expensive) political polls are not shared with the public because campaign staffs are planning tactics on their basis. These are termed internal polls. If you see them and you're not a campaign wonk, they're bogus.
I do not intend this to disparage polling at all - it's simply that like most things in politics it isn't quite what it appears to be. All IMHO, of course.
So true. What do you think of Nate Silver, since he does the latter?