I’m wondering about all the political polls that keep coming out.
Does anyone wonder about their accuracy? Maybe I’m being stupid. But I never answer my landline unless I have to. H gets calls from his work and my mom calls it buts that’s about it. My kids don’t even have landlines and wouldn’t answer their phone unless they knew the caller. So how can polls be accurate when no one answers or even has landlines anymore.
I guess I wonder when Donald Trump was suppose to be way ahead in the polls but Ted Cruz ended up winning. I just wonder if pollsters are going the way of the fax machine lol!
I think much depend upon whom is willing to answer the phone. We refuse to take polls, with the polls making us wonder if the world had gone crazy. Perhaps not as much as we thought.
I’ve heard that it’s much more work to get accurate poll numbers now. We have a landline still, but never answer it. FiveThirtyEight is a blog I guess, so no links, but they have a rating of polls there.
They also say that caucuses are harder to predict than secret ballots and that the more people running the harder it is as tactical voting may Trump (jk) who the voter actually likes best.
I did polling work way back in graduate school ( late 1980’s) and then they were amazingly accurate. I think they are less so now but not by much. I’m sure the polling firms factor in use of cell phones now. People generally don’t understand how a poll of 500 or 1000 people can determine outcomes, but many factors go into the selection of who is polled.
The polls in the last presidential election were also accurate. People got confused because certain people were saying the polls were “skewed” but in fact they weren’t. The Romney camp believed they were skewed too. Oops.
You can learn 50 times more than you imagined you might want to know about polls and their accuracy at FiveThirtyEight.com. Nate Silver, the founder of FiveThirtyEight, with his method of combining and weighting separate polls, pretty much ran the table in predicting outcomes in 2008 and 2010, and didn’t do much worse in the last two national elections.
I still answer the landline but we have it set up so I know whether the person is calling from a boiler room. It only rings once and then stops. Saves me a lot of running to the phone. If a polling outfit got through the screening system, I still wouldn’t answer. Or I’d answer “None of your business”.
I second (or third) the recommendation for Nate Silver and FiveThirtyEight.com. Not only are his analyses eerily accurate, but he writes and explains well.
His blog covers politics, sports,economics, and science.
In 2012, there were at least three sites that aggregated state polls (the ones that matter for determining who gets electoral college votes). All of them correctly predicted the election result and at least 47 or 48 states. Editorial commentary on those sites indicated that two were D-leaning and one was R-leaning.
Some polls are more accurate than others, the science behind the polling itself is pretty accurate. The real issue these days is actually getting people to respond to polls, in the days of landlines and before most people had things like caller id and voicemail, it was a lot easier to get people to respond. I don’t think how they poll has quite caught up to the new reality, and it is why some polls are hard to judge.
Too, there also is a bias in polling, in that often those most willing to answer a poll are those most hardened in their views. So as a hypothetical example, polls might show that candidate X is the one most people will vote for, when in reality it could be that people who support candidate X are more likely to want to take a poll, whereas someone whose opinions are not so informed, where they aren’t as politically motivated, may likely say no.
“In 2012, there were at least three sites that aggregated state polls (the ones that matter for determining who gets electoral college votes). All of them correctly predicted the election result and at least 47 or 48 states. Editorial commentary on those sites indicated that two were D-leaning and one was R-leaning.”
Polling entities don’t lean D or R. A poll can appear to lean one way or another but that is because of the nature of where the poll is being conducted. A poll needs to take into account the number of registered D’s R’s and I’s in a state being polled. Where D’s outnumber R’s a pollster will want more D’s in the poll and vice versa. That is why so many people in 2012 thought the polls were skewed - they believed it was because not as many R’s were being polled. It went over their heads that there are more registered D’s. There were also a lot of people who didnt understand the difference in voting patterns between an off year election and a presidential year and believed D’s were being over sampled in 2014. So they “unskewed” the polls and believed the R candidate was going to win. Even the pundits were doing this which boggled my mind.
That bring said, the polls account for bias and other factors which is why they all have a + and - margin of error.
People love to believe polls are flawed because of a whole assortment of things - especially when the say “I’ve never been polled,” but very small numbers are all that is ever needed.
The three web sites in question aggregated non-partisan polls in an unbiased fashion. However, their editorial comments made it clear that the writers had political preferences (2 D, 1 R), even as they reported their non-partisan poll aggregations in an unbiased fashion (resulting in essentially the same predictions for all three).
I have nearly 30 years of experience in market research. I could go on this topic for hours, but suffice to say that FiveThirtyEight is a very good source, and I second the suggestion to poke around there.
Personally, I haven’t used phone polling in years due to the issues identified above - landlines vs cells, who answers “unknown” calls any more, and so forth. I do almost all of my research online, using pre-identified panels because they are inherently less biased and I can balance them demographically.
So back to the question at hand, eventually landlines are going away, so how will polls be conducted? No one wants to answer a pollster on a cell phone, although I am a gallup member and I do answer. Online polls just dont cut it.
Online polls can cut it if they aren’t self-elected / opt-in. It’s like when you see a poll that says “Is Obama helping or hurting the country?” only those people who feel passionately one way or the other answer - the answer isn’t representative of the population at large.
And I particularly hate when news media trot out “focus groups” where they pick 8 people in Iowa or whatever. They are truly amateurish at it – you use them to gain insight into hypotheses, emotions, attitudes, beliefs and practices, but only a fool says “I’ve talked to 8 women in Des Moines and now I know …” They make real, true market researchers cringe with how they misuse the tool.
Enough people will be called on their cell phones and enough will answer and agree to be polled to conduct polls. Just because “you” won’t doesn’t mean there won’t be enough who will. In my experience people love being asked their opinion on things. This board is a perfect example of that!
