Re: GA senatorial campaign
Originally Posted by
Saab2000
Question regarding polling:
Because I’ve taken a bit more of an active role in recent years being on a mailing list of a local party branch (though not a member), I get a lot of texts with links to polling. I delete the messages because I’m skeptical of them. I know my phone number and email address are sold.
It used to be that polls were fairly accurate though imperfect. Nowadays most landlines are gone and most folks screen their calls far more thoroughly. We saw in 2016 that polling data in the age of the smartphone and the internet is less reliable than ever.
Thoughts on this and how reliable data is these days? Anyone know how polling data is acquired in 2022?
Way back in the 2005-ish time, there started to be a huge disconnect not only in political polling but all market research/ tv ratings etc about "cell phone only homes" because all kinds of stuff was getting way out of whack. Also, the way research has always been done it is trying to measure people by type (for instance Caucasian, male, 35-49 years old, college educated, professional managerial, lives in an A sized county, divorced, home owner, and about a zillion other attributes). What was happening as cellphone usage grew it was affecting different segments of the population differently. So that put another total fudge factor onto everything.
Ironically, now that cellphones are ubiquitous that is lessened. Also, not only ironically but distressingly to those who worry ethically about what they do with data, because of the data available with the UI's in cellphones and online internet tracking it can turn a problem into opportunity. Although the opportunity can be scary ethically.
Now, the discussion is how does the industry move from "looks like" in other words matching all of those segments and then extrapolating to the population correcting for under/over sampling of given segments to a way to do a better "acts like". There are some 25 year olds who act like 70 year olds. And how do you explain the 80 year old dads who are thrilled when their granddaughter gives them tickets to a Dua Lipa concert.
And then you also have sample bias and misreporting...simplistically although apropos of your OP...in surveys about TV everyone says they watch PBS but in truth they are streaming internet porn from their computer to the TV. All of that has to be corrected for (even examples not as extreme). In political polling it has long been known that folks tend to report what they think the person on the other end of the line (or even a machine) WANTS to hear. That is why extremes/ racism/ sexual proclivities/ etc so often make for polling that is way less than accurate.
All of this increase sample error and deviations etc. as adjustments are made.
All of the above is to give a bit more detail (prolly too much) of the goings on in what @JoB posted.
It is why data science has become more akin to anthropology than biology of late.
Last edited by htwoopup; 10-08-2022 at 10:02 AM.
Reason: Typos, typos typos
« If I knew what I was doing, I’d be doing it right now »
-Jon Mandel
Bookmarks