Election polling increasingly unreliable - We can hope for a Hillary landslide

[womanofthehills]

cellphone problems - cellphones must be dialed which costs more and you are not sure the person lives in the state you are surveying - people not answering their phones - response rate under 10 percent.

Election polling is in near crisis, and we pollsters know. Two trends are driving the increasing unreliability of election and other polling in the United States: the growth of cellphones and the decline in people willing to answer surveys. Coupled, they have made high-quality research much more expensive to do, so there is less of it. This has opened the door for less scientifically based, less well-tested techniques. To top it off, a perennial election polling problem, how to identify “likely voters,” has become even thornier.

Since cellphones generally have separate exchanges from landlines, statisticians have solved the problem of finding them for our samples by using what we call “dual sampling frames” — separate random samples of cell and landline exchanges. The problem is that the 1991 Telephone Consumer Protection Act has been interpreted by the Federal Communications Commission to prohibit the calling of cellphones through automatic dialers, in which calls are passed to live interviewers only after a person picks up the phone. To complete a 1,000-person survey, it’s not unusual to have to dial more than 20,000 random numbers, most of which do not go to actual working telephone numbers. Dialing manually for cellphones takes a great deal of paid interviewer time, and pollsters also compensate cellphone respondents with as much as $10 for their lost minutes.

http://www.nytimes.com/2015/06/21/opinion/sunday/whats-the-matter-with-polling.html?_r=0

and from PewResearch

Many people have a cellphone number with an area code that does not represent where they actually live. How do you know you’re getting a good sample if the location of a person does not match their cellphone?

What you’re describing is a phenomenon known as “under-coverage” and “over-coverage.” This isn’t an issue in national polling, since virtually every adult we reach in the U.S. is eligible for the survey, regardless of what location we thought we were calling. However, for state and local polling, under- and over-coverage can be a big issue. For example, in a recent national poll, 8% of people interviewed by cellphone in California had a phone number from a state other than California. Similarly, of the people called on a cellphone number associated with California, 10% were interviewed in a different state.

In a previous report, we discussed ways that researchers can attempt to correct this problem by attempting to merge in the commercially available billing ZIP code or full address associated with the cellphone number, though this is not available for all cellphone numbers.

Why don’t pollsters include more cellphones in their surveys?

The biggest reason is cost. According to federal regulations, cellphones have to be manually dialed by an interviewer, whereas landlines can be dialed automatically using a device known as an autodialer. Manually dialing cellphone numbers takes time, which increases interviewing costs. Each cellphone interview can cost almost twice as much as each landline interview. For this reason, some pollsters choose to either dial fewer cellphones or to exclude them from their sample altogether.
http://www.pewresearch.org/fact-tank/2016/01/05/pew-research-center-will-call-75-cellphones-for-surveys-in-2016/

[Foggyhill]

1. Its still an issue in national polling since to have a random unbiased coverage

you have to actually poll uniformly across states, rural, suburban and city boundaries.
Cell phone numbers don't tell you were those things start and end.
The result is basically a non random sampling of the underlying pop.
If the poll is small enough in number, this can easily introduce a quite substantial unknown error.

If the sampling is not a random sample of likely voters, then statistical analysis of the result is pretty iffy.

The fact you have to call so many people to get answers lead to another issue, self selecting.
The people most likely to answer those calls don't have the same demo as the overall population of likely voters. Tends to be whiter and older. Same thing for land lines; whiter and older.

As a whole, polling is more and more of a mess.

National polls should be a minimum of 1800-2000 people, anything under that is ludicrous.

[womanofthehills]

2. Agreed - most people I know are very busy and screen all their calls

the whiter and older are less likely to do that.

My landline talks to me and tells me who is on the phone and I get virtually no solicitations on my cell.

[Dem2]

3. Good info, thanks

I can only shake my head and think that the only real solution is to perfect internet polling.

[Demsrule86]

4. I lived in WI and OH most recently but still have my Georgia phone numbers on

my cell.

[Island Blue]

5. Even with landlines

I would hazard to guess that most folks don't answer the phone if it's a number they don't recognize. I know I don't.

[Tess49]

7. I don't either. n/t

[napi21]

6. This sounds like a real problem for all pollsters, nationwide.

Nobody has any idea how much of an impact the poll results thus none of the polls are credible.

I hope that means only a few hundred thousand LOUD nuts are REALLY voting for the CON and Tuesday will be a landslide for Hillary and will include the Senate & the House!