10-16-2017, 05:50 PM
(10-16-2017, 05:26 PM)hollodero Wrote: Is that so. Could you answer me two questions then, a) who are those people actually giving answers to pollsters and how can they possibly be representative and b) did you also foresee our green party falling out of parliament.
With our polls, maybe there was some distinct quality control going on after recent incidents that can only be described as embarrassing prediction debacles.
A: No, I didn't see that coming, though it would've been tough based on polling. PILZ, NEOS, and GRÜNE were all polling similarly. NEOS and PILZ ended up with results not to far off what the average polling said, but GRÜNE underperformed. I have a feeling this may be a result of some voters moving to parties not listed in some polls. Without going through each and every poll, though, that'd be hard to say for certain. I would've expected PILZ to be the underperforming one, given the new kid on the block status. Regardless of that, I think the results were well within the margins for all of them.
B: Each poll is different, but when we poll we strive for a representative sample. You start off with a large number of people, which gets smaller. Pew sees about a 9% response rate for their polls, for instance. That 9% may be several thousand people, though. Then you use demographic questions to categorize those responding to the questions. Once you have them categorized, you use the responses themselves to come up with the statistics. What you see as the margin of error (the +/- amount) is a calculation that is based upon how representative the sample is and how the respondents in those categories answered. So the more a sample deviates from the population, the greater the margin of error.
When I look at polls, I look at longstanding, reputable firms. I also, though, like to look at aggregated polling averages. What this does is help to even out some of the inherent biases, whether it be selection bias, bias in the wording of the question, or maybe what questions they ask at all. When you look at these averages of multiple polls, it gives you a great picture of what the public opinion is on a topic. It's essentially looking at the statistics of statistics of public polling, which gets a little meta, but it's fun. In the field I study in, we use this to gauge how politically viable a policy idea may be. We can write up a policy proposal that is heavily evidence based and is air tight in that it will benefit society in a large number of ways, but if public opinion is against it then it has to be shelved until the environment is more friendly and it can become more viable in the political arena.
