How has current telephone habits affected polling?

Back in the day everyone had a simple phone with no caller id or answering machine. When it rang, you answered it. Mostly it would be someone you know and occasionally it would be a survey or something. We were not yet inundated with telemarketers or scammers. But now, most people have caller id and answering machines. If they don’t recognize a number, they don’t pick up. How is this affecting the accuracy of telephone polls?

Occasionally I have answered an unknown call and it’s been a survey, but I’m leery of answering because I wonder about the motivations of the person. Is it an objective poll meant to gauge opinion? Or is it some slimeball who’s going to twist my answers to show support for his pet cause? Or maybe it’s some scammer who wants to get a personality profile on me for more targetted telemarketing calls. Usually I decline to participate because I assume the person is a scammer.

So it seems that the people who participate in the polling calls are people who answer random numbers and feel comfortable answering those sorts of questions. That seems like a small subset of the actual population and their opinion will be oversampled. Won’t this affect the accuracy of the polls?

I just got a call from a number I didn’t recognize. Said hello, greeted by silence. After a delay while the dialing computer routed the call to the phone-bank operator, I heard “Hello?”, which was my cue to hang up. Could have been a telemarketer, could have been a survey…don’t know, don’t care.

I haven’t had a land-line for at least ten years. My gf has one, but only answers if it is a relative (friends/work call her cell). The phone registers five or more calls a day. She gets maybe three a week that she answers.

The story goes that “Dewey Defeats Truman” was a result of the very first telephone polling. Truman did a lot of grassroot campaigning out in the rural areas where fewer people had phones, so th polling results were skewed. So it’s nothing new…

The biggest effect, currently, is that many people now only have a cell phone, with no landline. It’s illegal under most circumstances to call cell phones with a recording, so many polling firms don’t bother with them, but cell phone use is correlated with political views, which can introduce a skew into the poll results.

I actually purchased a machine to screen out the pollsters - they would *not *take no for an answer. They were so annoying, I thought about giving them false answers, because they pissed me off. Which makes me wonder about the accuracy as well - I mean, I can’t be the *only *one that these people have pissed off. (I never did it, now my device screens them out and I don’t even hear the phone ring.)

That may have been true, but the main thing there was that there was a high undecided response and most of it went to Truman. By 1948 virtually everyone had a phone. But in the depths of the depression, in 1936, telephones were much less common. A magazine called Literary Digest polled, I believe, a million people(!) with telephones and predicted Landon by a landslide. Landon actually won just two states (Maine and Vermont). And the Literary Digest went out business. This was always described as cause and effect, but I wonder if just a missed election prediction could drive a successful magazine out of business.

Anyway, pollsters learned from that fiasco that it wasn’t the numbers that were important, it was getting a reliable sample. I think they use just a few thousand, but work hard to get a random sample. That is why I think the OP has an excellent point. There is no reason to think that the people who actually answer their phones are a random sample of the population and I have no idea how they ensure the randomness of their samples.

You know what they call people who don’t answer their phones?

Democrats. :slight_smile:

One group that I think would be much more likely to respond to telephone polls are the elderly. As a group, I would think they are much more likely to have a landline, less likely to have caller id, and more likely to answer every call and respond to the questioner.

You’re thinking about the The Literary Digest Poll of 1936, which predicted a landslide for Alf Landon. The myth was that it was a telephone poll, but, in reality, it wasn’t.

The magazine did a mail poll, asking its subscribers to mail in their choices. It had worked for them in the past, but the problem was that in 1932, Roosevelt supporters – even those with phones, cars, and Literary Digest subscriptions – didn’t bother to return to survey forms. This biased the response.

On the other hand, the elderly are also more likely to vote in the actual election, so this bias is at least somewhat canceled out (assuming that you want the poll to reflect the people who are actually voting, as opposed to the population as a whole).

I don’t live in the US, but I’ve gotten a lot of calls from pollsters over the years (to my landline only). And after asking the questions about the election, they ask me for a bit of personal data: are you over 30 (or 50) yrs old", “the average salary is xxx–do you make more than that or less?” , how many years of schooling do you have.?

Throughout the conversation, I can hear the keyboard clicking as the pollster types my answers
I assume that the computer waits to see my personal data, and then adds my political answers to the data base only if the pollsters have not yet filled the quota for people, say, over 50, average paid, with over 12 years of schooling, etc.

They don’t use quotas, since there’s no sense in throwing perfectly good data away: Even if they already have plenty of people with your demographics, it never hurts to beat down the error bars a bit more. They do weight the results according to demographics, though.

I read that only 9 percent of people respond to telephone polls, down from 36% fifteen years ago. This means that more calls have to be made in order to get a representative sample. Also since there is more of a chance that the respondents are not representative of voters, more weighting of polls have to be done. The weighting of polls can introduce errors depending on the assumptions the pollsters make.