Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
Another Reason to Distrust Polls. This Time, It’s Evidence Based.
We all get those annoying headlines stating “Biden is ahead of Trump by X.” And we’re told that it’s based on science and Big Data, so you can take it to the bank. Well, any amount of time around conservative sites like Ricochet knows that we try very hard not to take “Biden is ahead of Trump by X” as gospel.
Today, we have a legitimate, authoritative, reason to distrust polls. Today, the US Supreme Court announced its decision in Barr v American Association of Political Consultants Inc. For a little background, the Telephone Consumer Protection Act of 1991 prohibits companies from making robocalls to cellphones. Congress, in 2015, decided to add an exception to the TCPA for purposes of collecting on any debt owed to or guaranteed by the U.S. Government.
The American Association of Political Consultants Inc challenged this exception as a violation of the First Amendment. In doing so, their hope was that the entire TCPA, one of the most popular laws passed by Congress in the last 30-40 years, would also be an unconstitutional violation of the First Amendment. The 4th Circuit, when presented with this lawsuit, agreed with the AAPC that the government debt exception was unconstitutional, but that the exception could be severed from the remainder of the TCPA.
The Supreme Court, in a 6-3 decision written by Justice Kavanaugh, agreed that the US Govt debt exception could be severed from the rest of the TCPA. This means the very popular ban on robocalls to cellphones remains in force and those folks from the American Association of Political Consultants Inc cannot robocall and force-feed push polls to your cellphone. Which means, to put it kindly, a whole lot of what pollsters and political consultants measure when it comes to polling is, at best, incomplete.
So we get doubts about the efficacy of polls in addition to not having our cellphones subject to the scourge of political robocalls and push polls. A pretty good day, I would say.
Published in General
Since I retired I mainly use our home landline.
I just recently upgraded from Morse telegraphy because this voice thing seems promising. And I got so much less political junk mail through the Pony Express.
For our Atlantic Telephone and Telegraph Service (ATT for you youngsters) it turns out that a free service NOMOROBO has proved remarkably effective at reducing political calls. We get one ring and then poof, the call disappears, routed to another number at NOMOROBO. I’m reliably informed that at NOMOROBO, Satan himself answers the political call, and the brimstone smell from the phone is so off-putting to the pollster that he assumes he’s dialed a Democrat, hangs up, and records yet another Biden voter.
Yaay! Good news!
Someone needs to do data on how many families only have cell phones. The trend started as early as 2003 I think, and has only gotten more common.
I live near a nuclear plant, so I have kept my land line. I know that is probably irrational. And there are days when I wonder if keeping it silly, but only a few.
We use NOMOROBO on our cellphones as well. I think it’s a couple of bucks a month – WELL WORTH it.
Somehow I don’t think this will stop them. Human beings are clever, and try to find ways around obstacles if they can’t run over them . . .
This states the US has been majority cellphone only since at least 2015, per the date on the link. And since at least 2017, only 6.5% of households have only a landline. Care to guess the median age of those who are landline only?
You would get self-selection bias either way. This just lets you not have to say no.
There were other reasons before that. Trump having 90% R approval and Rs being roughly 50% of the population, any poll that puts Trump’s approval rating under 40% is lying through their teeth.
At one point, a poll was published that between Biden and Trump among likely voters, Trump only gets 38% of the vote. I said that’s more accurate if you add “Democrat” after “likely”
55?
The implication is that those who are Landline Only tend to be quite elderly. My guess is that the median age is at least 70. So those pollsters and political consultants may already not be getting their desired information.
Another thing to also consider: Per this study from, of all agencies, the CDC, comes the following paragraph:
“Geographic differences were also
noted. Adults living in the Midwest
(53.0%), South (55.5%), and West (53.4%)
were more likely than those living in the
Northeast (34.2%) to be living in
households with only wireless telephones.”
This may give us even more insight as to the part of the country the pollsters call the most. After all, when more than 50% of US Households are cellphone only, and that most part of the country outside the Northeast are cellphone only, it makes one question how much pollsters and political consultants are actually missing when they take their polls.
Spring for a phone with the first ring suppression option and that one ring is also silenced. Now, if only NOMOROBO could just get a better handle on local exchange spoofing . . .
Yeah, polls-schmolls. However, you’d think the betting markets would account for that, yet they show a sizeable likelihood of a Biden win. Why?
ANTICIPATING OBJECTIONS: The betting markets also showed a sizeable likelihood of a Clinton win — as late as 9:00 pm EST on election night, 2016. Anybody have a theory why? (Note that “Democrats cornering the market” is probably not a good explanation, since it is the marginal bettor whose estimate matters.) Maybe the betting market was just wrong, or maybe the unlikely thing happened — the odds weren’t 100:1, after all. But if the polls are wrong, shouldn’t people putting their money where their mouths are know that? Especially given what happened last time?
Yes, we get a lot of local exchange spoofing here. We don’t pick up any call we don’t recognize and aren’t expecting. Mrs Rodin’s 90 year old mother tries to answer every call on her phone (usually unsuccessfully as she fumbles through her purse — the equivalent of a homeless person’s shopping cart) even though I tell her we can always check where the call came from if they don’t leave a message. When we moved her in to our home 2 years ago we had her landline number transferred to her cell phone as a means of making it easy for people who knew her home phone to stay in touch.
Distrust of polls is entirely appropriate, but you should have the same skepticism for the “polls are inaccurate!” cries of those trailing in the polls.
???
Rs are not 50% of the population – its down in the 30s somewhere. Same with Ds.
I don’t know. I really don’t do survey or polling methodology, but I talk to people who do (though it isn’t often about polling). Many of these people are clever and concerned with their reputations and their methods (sometimes weighting results really is just a secret sauce/intuition thing). I say that because they think about this stuff and will adjust their methods accordingly. So… Yes, I tend to believe that Trump is a deeply unpopular president (measured nationally) and he is likely to lose. The polling has been consistent.
If there is a problem with polling, it strikes me that the issue really isn’t the methodology of pollsters– it’s 1) changing samples between polls and 2) most American’s polling and survey statements are contaminated by all sorts of response biases and non-attitudes. That’s a harder subject to get around because you often won’t be able to ask the questions you’d like to ask without a lot of pre-tests, and that’s just not how modern “Who would you vote for” style polling works. But when you aren’t crunching out these weekly polls, and decide to put together something more rigorous (and expensive), the methodology of polling is fairly sound.
Again, I’m only questioning polls based on the pollsters only being legally able to gather surveys from limited sources (i.e. landline phones). BTW, how much money changes hands in those betting markets, and are they US based?
There’s also something to the theory that Trump “underpolls” because a good number of those who vote for him do not acknowledge that they are voting for him.
Understood. My challenge * is that people who are putting up money seem to agree with the polls. So maybe they’re “right” despite their flaws.
The Stossel-Maxim link I put in my comments says that almost $60 million have been bet. I believe they are illegal in the US.
* I take that back. Actually my point is: why were the betting markets so wrong in 2016??
I remember years ago Jay Cost talking about how before polling Poly Sci people would write great big books on politics and understanding the voters of the time and place and those guides were used to figure out what voters want.
We will likely see a return to that, and away from the constant use of polls to guide decisions. I think that will be for the betterment of everyone.
Also everyone is now going to go to focus groups.
Remember CNN used to have focus groups…Notice they dont have them anymore…
A good number of those who vote for him have no time to participate in polls and hang up instead.
$60 million dollars is…quite a small market for something that purports to allow betting on something as important as a US Presidential election. And as far as 2016, my guess is that the illegal-in-the-US betting markets were going off available poll averages found on RealClearPolitics and other sources. I seriously doubt that small of a market would justify independent spending on intelligence gathering beyond poll averages.
I presume you mean going away from focus groups, right? Focus groups and heavy polling during the last 20-30 years have often gone hand-in-hand.
I have no idea about the context of the statement. But generally I’m extremely dubious that polls and surveys will be used less. For better or worse political behavior will be survey and experiment based. I think far more attention is now being paid to desiring repeated surveys with the same respondents (good!) and there is a movement against taking “reported votes” seriously (a lot of people lie and say they voted for the whoever is president). But— unless you are a top tier person you won’t get away with “I talked to voters and this what I learned” as your main analysis, whether it is a book or an article.
(I say this and yes– really bad stuff has been published.)
Maybe. But it’s big enough that it would cost a lot to corner it if, for example, you thought having your candidate listed as the favorite was an advantage. I’m not sure how to calculate the cost, but you’d have to be willing to take every bid, whether from a Trump supporter or just from somebody who saw a profit-making opportunity. Would that benefit be the best way to spend your money if you wanted a Biden win? I doubt it.
I have this conversation about sports betting sometimes. “Notre Dame has so many fans, so the odds-makers have to give points to their opponents to equalize the betting.” Nope. If that were true, you’d get a flood of bets from “smart money,” which would quickly put it right. The bets from fans are inframarginal — irrelevant. So it is, I would think, for political fans.
Your theory is that the betting market just reflects the polling averages (or maybe the 538 probability calculation)? I guess that’s as good a theory as any. Going back to the sports analogy, though, people who have systems to calculate who will win a game are often careful to say that they do not expect their calculation to beat Vegas over the long run — especially if their calculation has success, in which case it will become part of the publicly available information priced into future betting lines. Even for games that aren’t the Super Bowl.
I don’t get it.
Also every poll has a target model. That target itself is a guess on how the voters will break down into categories in voting the next cycle.
So a poll that decides, for example that 30% of future voters will be Republican, and 35% will be Independent and 35% Democrat, does not ‘poll’ for which type of these categories are more likely to agree to being polled in the first place.
I strongly suspect that Democrats are more amenable to responding to polls. I also suspect that strong Trump supporters are split on responding to polls. Some, of course, want to talk to pollsters and proclaim their support. Others however are reluctant. I imagine independents who support Trump are also split in their response enthusiasm.
Cut to …filling-out the prescribed model. Easier to fill quotas for the Democrat 35%, of the Republicans and independents the respondents will tend to skew anti-Trump. In other words, they are getting a subset of a subset.
I write this having worked for three different polling companies conducting surveys: Gallup, Penn and Shoen, and Opinion Research.
Working at these companies (in the 1980’s) we always had to stay late to finish the quota for the category of Republicans.
If you think this is a tortured rationale, you haven’t explored polling models.
Very interesting. Thanks.