He specifically said they don’t do this.
He said they ask party affiliation at the end of the poll questions and their responses are for informational purposes only and they don’t do any weighting by party affiliation.
He we very specific.
That’s what I thought I was saying; but I can see that I should have been clearer about it. Let me try again.
If Republicans are systematically less likely to talk to pollsters — then there is selection bias (a major source of sample error)in the results. There are two common ways to help prevent selection bias from skewing results in a poll:
1. Asking for affiliation (or sex, or whatever other categories matter) first. If (say) the proportion of Republicans was less than expected (based on the known proportion of registered Republicans in the population), then registered Democrats would be skipped, and only Republicans interviewed, until the sample proportions matched the known proportions in the population. This would help make the sample more representative of the population.
This technique is commonly used, where the sex of the respondent is the important variable. Surveyors will follow some rule (e.g. interview a male, then a female, etc.). If the rule says they need to interview a female next, the interviewer will skip all male respondents, until a female is interviewed. That works well, because you expect a roughly 50/50 division between the sexes. In the case of political polling, things are probably too complicated (too many party affiliations, different percentages, etc.) That leaves the second method.
2. Ask for affiliation after all other questions, and use that information in the analysis of the survey. If (say) the proportion of registered Republicans sampled was smaller than the proportion in the population — then the samples would be weighted mathematically in the analysis.
Both of the above techniques are common in survey research. Newport said that Gallup doesn’t use the first method. Fair enough — but, they should then be using the second method. Otherwise, their results could suffer from selection bias. If Republicans are systematically less likely to talk to pollsters — then there is selection bias in the results.