Question about poll-creation policy

[baffie]

Can anyone here on DU tell me whether, when pollsters seek resondents to their polls, do they poll by percentage? More specifically, if there are, say, 60% likely Dem voters and 40% likely Repub voters in the area in question, do they poll 60% likely Dem voters, and 40% likely Repub voters?

I am always wondering, because, as far as I can see at this time, even if a poll is not deliberately skewed, if it is truly random it could easily contain 50-50, or some other variable which would not accurately represent the voting population.

[HFishbine]

1. It varies

Usually, this information is contained in footnotes, if at all. But it is an important consideration. Some pollsters take pains to use a sample that reflects the likely respondents accurately. Others do not.

As an example, look closely at polls about New Hampshire primaries. A poll of democrats is meaningless since over 30% of voters in the democratic primary in NH will likely be independents.

[Orangepeel]

2. It depends on the poll

most phone polls use random digit dialing to ensure a random sample. That is, they don't call from lists of people, they randomly choose numbers within representative area codes and hope they get someone. A computer does this, so a pollster doesn't have to wait on the phone until they get a hit.

They will probably then "qualify" the person who answers, asking them if they are a registered voter, a likely voter, etc. If it is random, they ought to get a sample that is representative of the party breakdowns, but, depending on the poll, they might make sure of this by asking first and then only giving the survey to people who qualify.