In response to some disappointing results in the Supreme Court this week, the President wondered aloud today on Twitter: "Do you get the impression that the Supreme Court doesn't like me?"
Sort of. A better way of putting this goes back to the Chief Justice's famous line in his confirmation hearing about being an umpire calling balls and strikes. As sports fans know all too well, though, star players gets more than their fair share of favorable calls. Perhaps that's because they command more respect from the referees. Perhaps it's because the referees think that they will get criticized more if they make a call against a star. Whatever the reason, that sort of bias is probably real.
A similar thought works for a president and the Supreme Court. Unpopular presidents fare worse in close cases that directly concern them. It's no coincidence that two of the leading decisions curbing presidential power (Youngstown and Nixon) came when President Truman and then President Nixon had approval ratings in the 20s. To some extent, this is just the Madisonian system: When one branch is weak the other branches try to take advantage. Strong presidents don't always win, of course, but they are more likely to win.
We won't know until the tax return cases are decided whether the Court is truly treating President Trump as a benchwarmer. But things are trending in that direction.
The DACA decision was pretty much a punt: yes you can do this, but you have to come up with a plausible justification. If Trump wins in Nov, DACA will be revoked. If the Rs hold the Senate, and Stephen Miller can do a slightly better job, the Court might even let that stick.
ReplyDeleteRoberts continues to play cute, presumably to stave off Dem calls to reform the Court.
"Multiple courts have told us this policy is unconstitutional without a statute, as did the former President who implemented it. Thus we're going to stop breaking the law."; How is that not a plausible justification?
ReplyDeleteI'm concerned that what we're looking at here is a modern "Switch in time"; Some of the conservatives on the Court have looked at the possibility of having mobs burn their homes to the ground, and the likelihood that everything they do is going to be overturned by a future Court, possibly a packed one, and have just thrown up their hands and given up.
Not worth risking their families over victories that will be shortlived.
Trump is feeling sorry for himself.
ReplyDeleteBostock had nothing to do with Trump. The argument over the meaning and application of the term "sex" in Title VII has been percolating for years.
In DHS v. Regents, Roberts once again demonstrated his personal crusade against extending any judicial deference to the bureaucracy, even if that aligns him occasionally with the progressive minority. When Roberts does this again to reverse a progressive policy priority, it will be fascinating to read the hypocritical dissents from the progressives signing off on the previous Roberts' opinions. Roberts only punted DACA rescission down the road and did no substantive damage to Trump's policy.
This is really baseless notion. I am pretty sure, that you haven't even read, and anyway understood the rulings.
ReplyDeleteSimple illustration, how baseless it is:
In the census case (Department of commerce v. NY) the chief justice, delivered the ruling, apparently, ruled against Trump, while the latter was at this best ( in terms of public opinion, before the Covid- 19 pandemic even, decided June, 27, 2019 ) and rejecting the agency reasoning for being unreasonable ( in administrative judicial review).
And what happened in this week:
In the DACA case, for the same reasons, the same chief justice, rejected the reasoning of the government, for being capricious, arbitrary, and in fact: post hoc rationalisation (means: explaining the action, after the fact, but, not explaining while the action taken).
So, in sum, in both cases, Roberts the chief justice, delivered the opinions, in both cases, ruled apparently against Trump. In both cases rejecting the government for the basically the same reasons, yet:
In the census case, Trump was good in public opinion, yet rejected, and now, bad, but rejected for the same reasoning in fact.
Just read first. Ah, forgotten, he(chief justice)considered conservative. Laughable simply.
Here to the census case at the time:
https://www.supremecourt.gov/opinions/18pdf/18-966_bq7c.pdf
Here to the current one(DACA):
https://www.supremecourt.gov/opinions/19pdf/18-587_5ifl.pdf
Thanks
P.S: I can bring so, countless cases. Maybe later....
el roam, your comment ignores a fundamental truth about the Trump Presidency. Only one day in his entire term did he have a positive net approval rating. Inauguration day. You'll remember that was the day with the biggest inauguration in history... or perhaps not, since that was his first of his thousands of lies and prevarications. This could be related to the Supreme Court questioning whether the explanations provided by his administration are "arbitrary and capricious." Just a thought, take it or leave it.
ReplyDeleteNever, not one day, did Donald Trump see a popular approval rating of even 50%. According to RCP, a conservative website, the highest approval was 47%.
This is an aberrant feature of this Presidency. No one has ever done this badly with public approval. Ford is probably closest in the modern era. Carter perhaps, but each had moments of popularity. Trump, by contrast, tries to be a populist, but has literally never been popular. Not one day. At least one thing which gives a president more informal power is popular support and he has never had this. Never. Underline it, circle it.
And forgotten:
ReplyDeleteBoth cases mentioned above, dealt with immigration issues. The utmost "cherished issue" of Trump, and conservatives as well.Yet, had nothing to do with nothing in relation to the rulings and Justice Roberts.
Unknown,
ReplyDeleteYou are right. But, of course I meant relatively, without simply mentioning it so.Because, according to the author of the post, there is significant fluctuation, which, influenced the decisions of the judges. So, that was the core of the issue. But, in terms of polls and public opinion and so forth... Your are right. It is all relative.Or rather, you are right in absolute terms( maybe) while not in relative terms( in this regard of the issue raised in the post).
"Some of the conservatives on the Court have looked at the possibility of having mobs burn their homes to the ground"
ReplyDeleteBircher Brett is an unmitigated, total wing-nut.
This is not a serious person.
I don't think this has anything to do with popularity. I actually agree with Bircher Bart that Bostock had nothing to do with that, it was the case that Scalia always talked about where textualism may or may not give a conservative victory. As to the Dreamer case, Roberts had given notice in the Census case that a half-butted effort might not pass muster, and from many accounts even favorable to the administration they brought not even a 1/4 butt to the table this time. It's a very minor rebuke to the administration.
ReplyDeleteI will add though that I do think the OP's point re sports and probably everything else is valid. If a largely unknown rookie in the basketball blocks the dunk of a superstar and the fall to the ground with no basket it seems to me that the ref is more likely to call a foul than not. But it's because they think: 'who is this guy? surely he did not block x's shot without fouling him?' To get over that the unknown player has to do a lot of demonstratively awesome things, then he will get the halo effect.
But this has nothing to do with the Dreamer case.
"Bircher Brett is an unmitigated, total wing-nut."
ReplyDeleteTo be fair, living in the Upside Down seems like it would be very scary.
This comment has been removed by the author.
ReplyDeleteUnknown: Trump, by contrast, tries to be a populist, but has literally never been popular.
ReplyDeleteThose who tell Dem media pollsters they "disapprove" of Trump or were "undecided" for whom they would vote were one of his largest voting blocks in 2016.
Have you ever considered they tell you what you want to hear to avoid receiving the same ration of sh_t you Democrats give Trump?
Our self appointed elites' "cancel culture" is reaching Chinese Cultural Revolution levels.
I'm wondering how those people polled know the media pollsters that call them are "Dem" folks. Or perhaps they just figure that all pollsters are "Dem" types. That would explain (if it were true, of course, which we doubt) the fact that even the pollsters Trump specifically hired to do "unbiased" polling told him he was unpopular.
ReplyDeleteThis kind of thinking is a ghastly mishmash of "No true scotsman" fallacies and epistemological closure. Sort of a Klein bottle of mental fixations. When wishful thinking is crossed with abject skepticism, the offspring is an odd beast.
I know it might seem naive, but figure that some law is mixed in here. Trump did so badly because his work was bad. Truman overshot as did Nixon.
ReplyDeleteTrump didn't lose that badly. The financial cases were slow walked (the tapes cases were rushed, decided in two months; the House in December said these cases were relevant to their impeachment power; case might be decided early July).
The Muslim Ban was upheld (it was after all Muslim Ban 3.0). The census business was decided on narrow grounds. They had a chance to try again; but apparently it was so obviously bad that they didn't. This too was decided on narrow grounds, too narrow probably, as Sotomayor noted. The lower courts restrained Trump some. He doesn't have too much complaint with the Supreme Court, really.
(Yes, lost the GLBT case, but such is the breaks with the type of pick Gorsuch is, textualist-wise. But, there is religious exemption loopholes even there. Plus, the law is clear there too really. It's not a matter of not liking Trump or something.)
Just to lighten things up, and support Gerald's idea, I am reminded of a story about a rookie pitcher facing Ted Williams.
ReplyDeleteThe umpire calls the first pitch, close, ball one, to which the pitcher objects mildly. The same with the second pitch, which draws a stronger response from the pitcher.
On the third pitch Williams hits a home run. While he is circling the bases the umpire steps toward the mound and says, "See, young man. When you throw a strike, Mr. Williams will let you know."
C2H5OH:
ReplyDeleteBoth the academia and the media who produce these polls overwhelmingly vote for, donate to and constantly propagandize for Democrats.
Their polling product during campaign cycles before the last couple weeks before the election overwhelmingly overstates support for Democrats. 538.com's collation of this propaganda polling has Biden up by 9 points over Trump.
If you believe this propaganda is objective reality, I will wager $10,000 here and now that Biden will not win by 9 points.
"Dem calls to reform the Court"
ReplyDeleteThis is a lie.
There are extremists within the Democratic coalition who want to pack the Court, which is not a "reform"- it's partisan stupidity.
"I know it might seem naive, but figure that some law is mixed in here."
ReplyDeleteLaw is ALWAYS mixed in here. There are rare exceptions, e.g., Bush v. Gore, but basically the Court works the way Karl Llewellyn said it would- the law is window dressing for their political positions, but it the window dressing is important because cases are not one-offs.
"On the third pitch Williams hits a home run. While he is circling the bases the umpire steps toward the mound and says, "See, young man. When you throw a strike, Mr. Williams will let you know.""
ReplyDeleteActually Bill Klem said that about Rogers Hornsby.
If you believe this propaganda is objective reality, I will wager $10,000 here and now that Biden will not win by 9 points.
ReplyDelete# posted by Blogger Bart DePalma : 10:43 PM
lol
These poll numbers are GREAT news for John McCain! - Sniffles DePalma (2008)
"but it the window dressing is important because cases are not one-offs."
ReplyDeleteI'll believe a lot of the Trump rulings aren't one-offs when we see similar rulings during a Democratic administration.
Bart, so you want to bet that Trump is not less popular than Carter in 1980? Because that's the last time a sitting president was defeated for reelection by such a margin.
ReplyDeleteSince Trump's approval rating is still a bit better than Carter's was in 1980, you are asking me to bet against the accuracy of the pollsters. Which seems to indicate that you secretly believe in the polls...
Here's what *I* think about Trump's popularity:
ReplyDeletePew: Trump’s approval ratings so far are unusually stable – and deeply partisan
Look at the general trend of the bottom line of that graph, approval of Presidents by members of the opposing party. With occasional blips and swings, it's been trending down for SEVENTY YEARS.
Partisan polarization has reached the point where the simple fact that Trump is a Republican guarantees that he will be disapproved of on every metric you might ask about, by effectively 100% of Democrats. Disapproving of Trump is now nothing more than a party identifier. If you're a Democrat, or affiliated with Democrats, you are automatically going to declare that Trump is stupid, smelly, not really a billionaire, bad on policy, evil, and doesn't floss after meals.
This means that he has an absolute ceiling on his approval rating of approximately 50%, because literally nothing he can do will make half of the population approve of him. He could commit sepiku in the Oval office, and Democrats would criticize him for getting blood on the wallpaper.
But, the electoral implications of this are limited, because virtually everyone who hates Trump was never going to vote for him anyway. At one time, because the parties were not so polarized, the only way a President could drop below 50% approval is if a significant fraction of his own base disliked him. And of course being unpopular with your own party was electoral poison!
But it's now possible to have a <50% popularity rating despite your own party being wildly enthused about you. Trump will be going into this election with the support of virtually every Republican voter. Something you couldn't have said about a Republican President with this low of overall popularity 20 years ago.
I don't think this helps him, if that expressed dislike by Democrats reflects actual intensity of opinion, rather than just a generalized, "I'm a Democrat, so I'm going to say he's horrible, because that's what Democrats believe." tendency. Obviously it would help him if Democrat leaning voters were more approving.
But the kiss of death it ain't.
Brett: I'll believe a lot of the Trump rulings aren't one-offs when we see similar rulings during a Democratic administration.
ReplyDeleteA flash from the not so distant past which Gerard also ignored: Obama Has Lost in the Supreme Court More Than Any Modern President
This comment has been removed by the author.
ReplyDeleteC2H5OH said...Bart, so you want to bet that Trump is not less popular than Carter in 1980?
ReplyDeleteYes. I will also bet $10,000 Trump will have a higher percentage of the vote in 2020 (barring death or disability) than the 41% of the vote Carter drew in 1980.
Once again, the only poll which matters in politics is the one where actual voters cast actual ballots.
Sure, Bart. And Trump is also not less popular than dead skunks, PG&E, and leprosy.
ReplyDeleteWhat these bets prove is that even Bart has reached the conclusion that Trump is massively unpopular. Probably John Roberts, who is far more intelligent than average, has also reached this conclusion...
C2H5OH:
ReplyDeleteYou and anonymous set the standards of "popularity." I just held you both to them and you naturally demurred.
Point proven.
Excuse me, but I think you mean "Unknown" -- and that individual pointed out that Trump's approval ratings have never approached anything normal. And your response was to cast all pollsters as "Dem" folks, who apparently are lied to by vast swathes of the populace. Which, I pointed out, cannot be believed.
ReplyDeleteAnd your response is to bet that Trump's approval (as measured in November) might be better than 41 percent?
I fail to see what you think that you have "proven" by making idiot bets, other than that you realize that, if you bet Trump might get close to 50, you would almost certainly lose.
"And Trump is also not less popular than dead skunks, PG&E, and leprosy."
ReplyDeleteYou left out syphilis. Though personally, I prefer syphilis.
You and anonymous set the standards of "popularity." I just held you both to them and you naturally demurred.
ReplyDeletePoint proven.
# posted by Blogger Bart DePalma : 9:31 AM
Was the point that you’re an idiot? If so, congrats! You “win”!
If the polls said something good for the GOP Bircher Bart would be touting them. Since they're not they're 'Dem polls.' He does the same with government experts or agencies, the media, etc., He's partisan incoherent.
ReplyDeleteThe conservative melt down over Bostock has been fun. Scalia and other textualists always said the approach would yeild politically liberal as well as conservative outcomes, that was supposed to show it wasn't results oriented. Of course what most conservatives were really after was results after all.
ReplyDeleteThe OP argues that when presidents (or whatever Trump is) are at a low point in their power that other institutions, including courts, are more likely to "fare better." Truman was cited. He was a Democrat (a member of the Democratic Party).
ReplyDeleteIt makes sense that Trump is not treated the same way as various others with his position because he is acting differently. That is why many voted for him. They supported (I'm right; I'm using italics; I should write Federalist Society briefs) him for such a reason as much as some did so for promoting alleged (cf. John Fea's book) evangelical values or a generally (which won't always mean they will be happy) a certain breed of judge, all things being equal.
One thing this week that upset some conservatives was the GLBT ruling. Well, that would also come down in a Democratic Administration. The Muslim Ban was upheld. Yes, we had two major administrative law decisions (census/DACA) where Trump lost, but it has been detailed how far the Trump Administration has pushed there. This includes a great number of attempts (often successful) to get higher courts to prematurely intervene in pending litigation. Even in the DACA case, only Sotomayor rejected a premature move to close off an equal protection argument EVEN BEING RAISED.
Anyway, just to note it, my "naive" comment was a bit tongue in cheek. The talk of pragmatic decision making here was often made by lawyers and law professors who are well aware of how law works. And, though some are more idealistic about it than politics, there is a mix of things involved. So, I wanted to flag that.
Mista:
ReplyDeleteBasically everyone who cares a ton about politics but doesn't practice law ends up caring more about results than law. Which is really annoying to lawyers who know how darned important predictability and stability in the legal system is.
You are right to call out the right, though, because they have been so obnoxious about their assertions that they are the only ones doing law and that everyone with a more liberal judicial philosophy is just legislating from the bench. That was never true but convinced a whole bunch of Republican non-lawyers.
The reality is that all the cures are worse than the disease. If you think SCOTUS is too political now, take a look at some of the states where politicians or the public have greater direct control of the courts, like Florida and Wisconsin. It's no comparison.
Now folks like Brett and Mark Field might prefer that, as long as their side gets the control the Court. But the people who will get screwed are all the average people whose disputes aren't discussed in political platform, and who will suddenly find that the law is far more unpredictable and their lawyers can no longer give them definitive advice. Political partisans, who think the only thing that matters is their side putting points on the board, don't care about those people.
Joe:
ReplyDeleteLaw professors didn't used to talk this way, because for the most part they talked to the public in law review articles and sometimes speeches and pieces not published in real time. Which meant sifting through the law was valued.
Now it's a race to get your hot take up on the Internet ASAP, especially Twitter. And that means there's no time to read cited cases or really figure out the legal basis for anything. The goal is to get a comment up. Which means Supreme Court Kremlinology has become the dominant mode of discourse about the court system.
No, they did law before and they do it now. Politics matter, but not in the all consuming matter that is portrayed on Twitter.
The conservative melt down over Bostock has been fun
ReplyDeleteI saw Bryan Garner, who was a friend and co-author of Scalia, said that he thought Scalia would have voted with the dissent here based on a sort of Kavanaugh "commonly understood" argument of some sort. It was also noted on the Strict Scrutiny Podcast that there was sort of a "two Scalia" situation -- he hardened into more of a, to be blunt, Fox News sort of judge as he got older.
The general sentiment stated by Mr. W. is correct as far as it goes though rather often there is an ideological overlap (in part since ideology is in some sense a mix of views that will include legal thought; also Scalia et. al. at times speak in cliches that in practice are a lot more complicated, in as they might admit if pressed). I also would note though (see Dorf on Law discussions on the case) that in practice textualism can be too simplistic without usage of things like purpose etc.
A common cite here is the Scalia opinion of Oncale v. Sundowner Offshore Services, Inc.:
As some courts have observed, male-on-male sexual harassment in the workplace was assuredly not the principal evil Congress was concerned with when it enacted Title VII. But statutory prohibitions often go beyond the principal evil to cover reasonably comparable evils, and it is ultimately the provisions of our laws rather than the principal concerns of our legislators by which we are governed.
I think "Our holding that this includes sexual harassment must extend to sexual harassment of any kind that meets the statutory requirements" was not some "obvious" understanding in the 1960s, but the statute ratified had open-ended text that -- as with other congressional legislation -- gained wider reach on specifics over time.
It could have been more narrow, but that is not what they passed. See also, equal protection as compared to specifically denying voting by race etc.
One other point- numerous public commentators think that reading a few Supreme Court opinions makes you an expert. It doesn't. Anyone can read SCOTUS opinions. But there are literally people out there who make $1000 an hour analyzing these things.
ReplyDeleteNow some of those people are unable to say useful things publicly because of the conflict of material interest (see, e.g., Lisa Blatt's dumb quotes about Brett Kavanaugh).
But there are others who can, and do, especially older SCOTUS litigators and those who have taken academic jobs later on life. And they RARELY comment on the Kremlinology. Those are the voices that should be elevated.
If we listened to the right people, and not the loud generalists, folks would have a better notion of what goes on (and this is true for other legal specialties too).
"Law professors didn't used to talk this way, because for the most part they talked to the public in law review articles and sometimes speeches and pieces not published in real time. Which meant sifting through the law was valued."
ReplyDeleteThis is silly stuff. There were law professors etc. that talked like this before and people were critical of it. They talked in real time too. Not that waiting a few months and saying the same thing changes the basic point here.
Lawyers who were "cynical" or "realistic" or whatever of how law worked were around since there were lawyers. Judging is by humans. It is a collection of things. If it was done by Vulcans or robots (at least ones not programmed by humans, I guess), maybe it would be different.
BTW, I think Scalia actually shows the power of legal realism. The only consistent thread in his gay rights jurisprudence is that any claim that could only be made by a gay person had to lose, because Scalia thought the ostracism of gay people had deep historical roots and any "good" society would be homophobic.
ReplyDeleteOncale was different because a hetero male could suffer harassment.
Scalia would have clearly dissented in Bostock.
"One other point- numerous public commentators think that reading a few Supreme Court opinions makes you an expert."
ReplyDeleteThe average person being an expert is sorta a conceit in this country, akin to back when anyone who could read the Bible felt they could determine what it means.
Of course, this isn't true, and even your average lawyer or law professor are not too familiar with specific areas of law. But, people can comment, as some people comment on politics etc. w/o being some kind of expert.
Note that "reading a few Supreme Court opinions" is often not exactly all the people who comment do though some might do better to reading even a few more closely.
Joe:
ReplyDeleteDuncan Kennedy existed. But he was an outlier. Most professors in the "realist" traditions DID NOT argue that the law didn't matter. Indeed, almost all of the "law doesn't matter" folks even now are outsiders from other fields, in part because it's such a simplistic and wrong take on the law.
And literally NO appellate litigators think law doesn't matter. But what do we know? After all, some person on the Internet must know more than we do.
"But, people can comment, as some people comment on politics etc. w/o being some kind of expert."
ReplyDeleteThe problem isn't commenting, it is drowning out the experts.
I first noticed this in the 2000 election. There's probably 500 lawyers and law professors nationwide who do election law, a very narrow and technical specialty. And yet all the loud voices were from the same generalists who comment on every politically important SCOTUS decision. They never yielded the floor even though most of them knew nothing about the actual subject.
I get that this goes back to the Protestant Bible. It's a long American tradition. But it really harms legal discourse.
For instance it's important to know why Sotomayor lost 8-1 on the equal protection issue in the DACA case. And yet all I read yesterday was "there must have been a deal". No, you don't get to 8-1 through a deal. It was more of "you would have to overturn two hugely important cases, AAADC v Reno and Ashcroft v Iqbal, and limit a third, Washington v Davis, for Sotomayor's position to work. A good civil rights lawyer could have told you the claim was a very long shot. But they aren't the loud voices in the room.
For those who are interested, here is the deal with polling:
ReplyDelete(1) Pollsters cannot obtain random samples because of mass refusal to participate. Instead, pollsters demographically reweight the responses they do obtain.
(2) Similarly, a large cadre respond, but refuse to give their preferences or lie to the pollster.
(3) Adults misrepresented as Americans: Polls of anyone self identifying as an adult reweighted by census polled demographics. Even if accurate, 7% of the respondents would be non-citizens and about half do not or cannot vote.
(4) Registered Voters: : Polls of anyone self identifying as a RV reweighted by census polled demographics. The primary error here is the demographics of the population at large when compared with the actual electorate overcount the young and minorities - Dem leaning groups. As 538.com recently admitted, these polls overcount Dems by “a few points.”
(5) Likely Voters: This is where the pollster guesses what they think or want the demographics of the electorate to look like, then they reweight actual responses by these guesses. In 2016, the Dem media pollsters wrongly assumed Obama’s young and minority base would turn out for a white establishment Dem and the white working class would stay home as in 2012. They were wrong.
I am only looking at enthusiasm to vote and the demographics of the “undecided.” If Trump maintains his current giant advantage in enthusiasm and the demographics of the undecided are predominantly white working class and married as in 2016, get ready for the polls to be wrong again.
On point 1, to amplify, response rates have been down around the single digits for some time now in political polling. That just eviscerates the theoretical basis for thinking it's reliable anymore. Yet the pollsters keep reporting confidence intervals that are based on the number of people who respond, and ignoring that the mathematical theory behind polling says those confidence intervals are garbage unless response rates are high.
ReplyDeleteBut, of course, what can they do? They're like Wile E. Coyote, the cliff has vanished from under their feet, and all they can to is avoid looking down so that they don't fall. What are they going to do except ignore the obvious problem? Just find a different line of work?
Like most things he opines on, Bircher Bart knows very little about survey research. Remember, this guy guaranteed a Romney presidency.
ReplyDelete"That just eviscerates the theoretical basis for thinking it's reliable anymore. "
ReplyDeleteFor those with a laypersons' understanding of how it works. I mean, Bircher Brett really thinks the people that make a living doing this don't have ways to account for and deal with this. But then again his view of expertise has always been this way.
There's a reason everyone, including Trump and Republicans, and businesses and such, pay lots of money for survey research. It's not perfect, but it has a lot to tell us.
Of course, this is neither here nor there. If the polls were saying things to make Republicans excited or such our Birchers would be touting them. These are partisan incoherents.
But it's now possible to have a <50% popularity rating despite your own party being wildly enthused about you.
ReplyDeleteI agree. But, there is likely a big electoral impact on the difference between a 47% and 43% approval rating, and Trump hasn't been able to break through above 45%.
Brett, the response rate makes no difference, mathematically, as long as the respondents are, in fact, a random sample of the population. You seem not to understand that. And the mathematics behind the relationship between confidence intervals and the size of the sample has a proof. Your assertion that a low response rate makes confidence intervals meaningless lacks one...unless you are claiming, somehow, that the low response rate implies that the sample is not random.
ReplyDeleteIs that your claim?
And Bart's idea that one can predict elections based on turnout at rallies and wishful thinking about the "undecided" is blatant wishful thinking, not a basis for meaningful prediction. That may be why the serious pollsters, who live and die based on their accuracy, don't use it.
get ready for the polls to be wrong again
ReplyDeleteAt the national level, they did pretty well in 2016. Trump was supposed to lose the national popular vote by 3.5% points. He lost by 2% points. And that result was consistent with Clinton's slightly higher (but still bad) favorability rating. Winning the close states is why Trump prevailed.
Biden is now up by about 9%-points in the national popular vote, which is consistent with Trump's approval rating as compared to Biden's favorability rating. I have little doubt if the election were held today, Biden would win. Trump needs to either drive up his own ratings (unlikely) or drive down Biden's (possible). But don't fool yourself into believing the current polls are not bad news for Trump.
"For those with a laypersons' understanding of how it works."
ReplyDeleteIn the sense of an engineer who works with numbers every day, sure, it's a layman's understanding. But when your response rate is in the single digits, it is, IMHO, criminally deceptive to be citing confidence intervals that are only theoretically sound if response rates are high.
"Brett, the response rate makes no difference, mathematically, as long as the respondents are, in fact, a random sample of the population."
ReplyDeleteThat is, with all respect, kind of a silly thing to say. You know for an absolute fact that your respondents are different from the general population, highly atypical: They responded to your poll! You just hope like hell that responding isn't correlated with how people answer any of the questions you're asking.
"And the mathematics behind the relationship between confidence intervals and the size of the sample has a proof."
Which proof is only mathematically valid at 100% response rates! Do you genuinely not understand that?
it is, IMHO, criminally deceptive to be citing confidence intervals that are only theoretically sound if response rates are high.
ReplyDeleteConfidence intervals are not theoretically unsound if response rates are low. They are just simply wider.
I'm not really inclined to parse the poll data too closely though do note that supposedly 2016 was a "black swan" event so if the polls were wrong, it very well could be not really too informative. But, such things, like reliance on the CDC, is for some a selective citation at any rate.
ReplyDeleteI am careful with poll data at this point though overall agree that if one is for Trump, one rather not have it. Being careful ("it's June" etc.) is something many who don't support Trump are saying too. Also, those whose job it is to actual parse poll data have noted too (noting there isn't any "one" specific conclusion here) that looking at the data closely shows that Biden is doing better than Clinton was, even noting that people were assuming Clinton was a lock then as well.
This all is not really just you know because people who are not experts are reading things wrongly since they are not experts. I personally have specific knowledge about certain things more than the average person while they know certain things more than I. But, a basic understanding of things still can be formulated & on that level, people can be wrong.
This is all rather important since in this country we have some respect for the people having more of a role in governing than might be the case. We don't expect the average voter (NY has a primary next Tuesday) or juror or member of the militia or whatever to be some sort of expert on everything. But, some basic reasoning and understanding is helpful. As with everything else, a good mix of humility helps.
I suppose I should elaborate, because there's a sort of equivocation pollsters engage in on this point.
ReplyDeleteThere are two confidence intervals at stake here:
1) The confidence interval that tells you, if you ran the poll multiple times, how much variation in the results you'd get. That's the one with the theoretical basis.
2) The confidence interval that tells you, if you run the poll once, how much variation you'll get from the actual population being polled.
At 100% response rates, they are essentially one and the same. As response rates drop, they become different if responding is at all correlated to how you answer questions.
Now, C2, which of these confidence intervals should anyone really care about?
"Confidence intervals are not theoretically unsound if response rates are low. They are just simply wider."
ReplyDeleteThe actual confidence interval is wider. The one reported by the pollsters isn't, because they calculate it, (Simplifying here.) based on the number of people who responded as the sample size, and so the number they report isn't taking into account response rates.
It doesn't, because the sample size based confidence interval is the only one that you have a theoretical basis for calculating.
There are two confidence intervals at stake here
ReplyDeleteThere is only one confidence interval of interest: 95% (or pick your favorite number) of the time the true mean will fall inside the confidence interval.
the number they report isn't taking into account response rates
ReplyDeleteThat is because the confidence interval only depends on the sample size, not the response rate.
Like most things he opines on, Bircher Bart knows very little about survey research. Remember, this guy guaranteed a Romney presidency.
ReplyDeleteAnd notably, the same arguments were made about skewed polls wrt Romney.
I'm still waiting, Brett, for a link to the proof that the confidence interval depends on the response rate. Unless and until you can provide one, you are simply engaging in selective perception.
ReplyDeleteYour claim apparently is that responding to the poll makes you more likely to vote for or against something or someone. Maybe. But to claim, further, that the inaccuracy comes down in favor of your desired result appears to me be wishful thinking. It may, in the present case, mean that Trump is even less popular than the 40 percent the pollsters say...
The fact is that polls, with very rare exceptions, have been remarkably accurate over the history of polling. Even recent polling, except when those polls were taken in a rapidly-changing environment (the last couple of weeks before the election of 2016, for example.) Trump has been polling where he polls today, pretty consistently his entire polling history. What's going to change that?
If you want to discuss polling more generally, I have a lot of problems with the industry. First, there's way too many polls. Second, the sample sizes are way too small. Third, the polling industry, and certain journalists, over-promote the results of polls, especially polls far out from an election. Fourth, the national numbers that pollsters promote, because they have them and don't have state numbers all the time, are not really important given the electoral college. Fifth, the industry isn't honest about stuff like the refusal rate and the shift to online polls. Sixth, people lie to pollsters anyway (and honestly, they should- if, say, 40 percent of the public lied to pollsters and polls became completely useless, we'd be a better country in a number of ways).
ReplyDeleteBut the problem is none of that particularly skews the polls in a liberal direction. Quite the contrary, pollsters make their money by at least claiming they are accurate. If you have a pollster that claims throughout the year that its client is going to win by 10 points, and the client loses by 8, that pollster isn't going to get much work next cycle.
The notion that polls are biased against conservatives is just low grade conspiracy theory. And I say that despite thinking that the entire enterprise of political polling is grossly corrosive of democracy, full of dishonesty and commercial misrepresentation, and basically an evil endeavor.
And by the way, if I were advising either party's presidential candidate, I would tell the candidate to encourage all of their supporters to lie to the pollsters.
ReplyDeleteIf all the polls returned junk results, that would totally prevent the media from ever reporting you were down in the polls. It would force the media to cover the candidate on substance.
Dilan -- of course, you mean "lie to all the other pollsters" -- because candidates rely on polling to adjust their messages and take the pulse of their campaigns, and you would probably be fired by your boss at the request of all her other advisers...
ReplyDeleteThat is because the confidence interval only depends on the sample size, not the response rate.
ReplyDeleteTo clarify, my statement is correct for large populations because the binomial distribution (repeated trials with replacement - not the reality) is very close to the hypergeometric distribution (repeated trails without replacement - the reality). The normal approximation to the binomial distribution does not depend on the response rate, and works well for large populations. The normal approximation to the hypergeometric distribution does depend on the response rate, and is required for small populations. Therefore, for small populations, response rate matters.
Presidential polls have very large populations, so the response rate does not matter .
For small populations with small response rates, you are talking about small-sample statistics, which is always going to be dicey.
ReplyDelete"I'm still waiting, Brett, for a link to the proof that the confidence interval depends on the response rate."
ReplyDeleteI think you're just demonstrating that you didn't follow the proof, or maybe didn't understand what I wrote. The calculation of the confidence interval depends only on sample size, (OK, slightly on population size, that's where the DoF comes in.) but that calculation implicitly assumes a genuinely random sample.
If you sample 100 people from a population, and all 100 respond, assuming your choice of people to sample was genuinely random, the calculation of confidence interval is valid.
BUT, if you sample 10,000 people from a population, and 100 respond, that your original selection of 10,000 people was genuinely random avails you not, because you can't assume the 100 people out of 10,000 who responded are representative of the population. In fact, you know they're NOT representative in terms of inclination to respond. So, the proof fails to be valid anymore.
Now, if inclination to respond isn't correlated with the answers to the questions you're asking, you're still good. But, how do you know that it isn't? Usually, you just hope like heck that it isn't, because the only way you can tell if it is, is to take the people who didn't respond, and pester them and pester them until they do, and compare that result to the original result, and hope that pestering them didn't change their answers. Which is actually expensive, and to pull off validly requires a really large initial sample, multiplying the expense.
And still doesn't make the original proof, which assumes 100% response rates, valid anymore.
"Your claim apparently is that responding to the poll makes you more likely to vote for or against something or someone. Maybe. But to claim, further, that the inaccuracy comes down in favor of your desired result appears to me be wishful thinking."
Yeah, that would be wishful thinking if I did claim that. The only thing I'm claiming here is that pollsters are over-stating the reliability of their polls by reporting confidence intervals based on sample size, when they know that there are other potential sources of error, and that some of them are getting frighteningly large.
Pollsters tend to herd, and the herd sometimes overstates the Republicans' chances, sometimes the Democrats, because they keep over-correcting. So I don't really know that the polls are understating Trump's appeal. Might even be overstating it.
The only thing I do know is that I'd be stupid to rely on them.
Oh, I understood perfectly. But the problem is that you have no idea how the pollsters come up with random samples.
ReplyDeleteI'll point out, since you seem to have mathematical problems, that sampling 10,000 people randomly and getting 1000 random responses is mathematically indistinguishable from randomly sampling 1000 people and getting 100 percent response.
Perhaps you haven't noticed, but polls conducted by different pollsters all seem to be pretty much in agreement with each other (caveats about the polls being "experiments" in probability, so there will be outliers.) If, as you propose, "non-response" affects the result, we should expect to see that. We don't.
"I'll point out, since you seem to have mathematical problems, that sampling 10,000 people randomly and getting 1000 random responses is mathematically indistinguishable from randomly sampling 1000 people and getting 100 percent response."
ReplyDeleteYou wrote that right after I explained in detail exactly why that wasn't true.
"Perhaps you haven't noticed, but polls conducted by different pollsters all seem to be pretty much in agreement with each other"
Yes, that's known as "herding". It happens because polling isn't just an exercise in mathematics conducted independently by multiple people, but is about half math and half black art, and the artists get nervous if their results are too different from the other artists. So the polls start converging on each other, much to the annoyance of aggegators like Nate Silver. (Because herding reduces the utility of having more than one poll.)
"If, as you propose, "non-response" affects the result, we should expect to see that. We don't."
I'm pretty sure you meant, "we shouldn't expect to see that. We do."
So, you're assuming that non-response is correlated to whether you plan to vote for Trump, but it's sort of randomly correlated, where each different pollster is going to see a different correlation? How exactly do you imagine that works? If Pew calls, Trump voters don't respond, if Rasmussen calls Biden voters hang up? Even though neither, in my experience, identifies themselves?
C2H5OH: Brett, the response rate makes no difference, mathematically, as long as the respondents are, in fact, a random sample of the population.
ReplyDeleteResponse rates differ between demographics, so it is impossible for any pollster to obtain a random sample of the population. They all reweight actual responses by some measure of demographics. I have noted why these measures often do not resemble the electorate.
And Bart's idea that one can predict elections based on turnout at rallies and wishful thinking about the "undecided" is blatant wishful thinking, not a basis for meaningful prediction. That may be why the serious pollsters, who live and die based on their accuracy, don't use it.
Serious pollsters do consider self-reported enthusiasm in deciding whom they will consider as likely voters
The Democrat media pollsters you take "seriously" have not pegged an election in the swing states in several cycles.
In the latest fail in 2016, the only pollsters who pegged the election were Rasmussen nationally and, more importantly, a GOP firm called Magellan in the swing states. Magellan succeeded where the Dem media universally failed because they pushed the large cadre of "undecided" to disclose their preference and found a reservoir of Trump voters.
Re: "herding" -- let's be clear about what a poll result actually means. It means that, if another poll is taken, at the same time and from the same underlying population, then the same result will be obtained, within the confidence interval, with a certain (high) probability.
ReplyDeleteSo how does one distinguish "herding" from the mathematically-expected result? Obviously, in Bart's and Brett's case, by wishful thinking.
Bart, if by "pegged" you mean "exactly", no, they haven't (rolls eyes at the inability of Bart to understand what polls actually mean). But if you mean, came close to the actual result when the poll was taken immediately before the election, then, no you are egregiously wrong. Show us all the cases to the contrary, as a percentage of all the cases, and we'll see what the probabilities are that polling does, in fact, work.
Hey, guys -- as with any other mathematical theory, sampling theory is only an abstraction which is useful for modeling real-world behavior, and, as such, it can be in error on occasion. But it's still a great deal better than the fevered wishes you entertain that, somehow, Trump is anything but a hugely unpopular president.
"In the sense of an engineer who works with numbers every day, sure,"
ReplyDeleteOf course it's not so much a matter of numbers as it is of methods and techniques which laypersons are ignorant of. Political polling is, on average, consistently remarkable accurate (even in 2016 for President they were very close to being on target).
It's interesting to note the general dynamic.
ReplyDeleteThere are people trained in and with experience in polling. People, corporations, politicians, the armed forces, etc., virtually all of them, pay top dollars for their work.
A smart layperson would assume these pollsters must be delivering something valuable though they don't know how they do it.
People like our Birchers though assume they understand the topic and the trained and experienced people are doing it wrong.
It's the usual inductive logic fail our Birchers make in so many other areas.
And let's be clear: they're just doing this because they think the poll results are dispiriting for their side. If it were otherwise they'd be touting the polls (Bircher Bart does so with enthusiasm in this very thread!).
ReplyDeleteThe only thing going on with them is partisanship, everything else independent of that, such as professional expertise, standards, etc., is meaningless to the Bircher.
"So how does one distinguish "herding" from the mathematically-expected result?"
ReplyDeleteBy the results of multiple polls converging on the same result more than the sample sizes would justify, obviously.
And, notice that in Silver's example, they didn't just converge, the actual result wasn't even inside the confidence interval: They converged on the wrong answer!
Setting aside herding, the point I'm trying to get across here is that the theory of statistics really IS based on an assumption of high response rates. It can, if you're lucky, keep working with low rates, but only as a matter of luck, which is eventually going to run out.
Oh, and note that by "working" I don't mean that the polls tend to give the same result, I mean that they tend to give a GOOD result. You could easily get a case where competently executed polls would cluster only to the extent statistical theory would predict, but because response happened to correlate with what was being polled, the polls all clustered around the wrong value.
"A smart layperson would assume these pollsters must be delivering something valuable though they don't know how they do it."
ReplyDeleteFirst, I DO know how they do it.
Second, pollsters deliver two DIFFERENT things of value, but both of them are delivered to the people paying for the polls.
The open, acknowledged thing of value, is accurate results. This is the thing of value people in politics are paying for with "internal" polling that is used to guide campaign actions.
The dirty little secret thing of value, is results that influence public opinion. This is the thing of value that people who pay for polls are expecting from polls you don't have to pay to see.
Just another application of, "If you're not paying for the product, YOU are the product."
Roberts doesn't like to be treated like a rubber stamp. He'll approve anything the admin wants if it's done the right way but he doesn't take to being told he has to believe obvious, stupid lies.
ReplyDeleteC2H5OH:
ReplyDeleteHerding refers to when Democrat media pollster results suddenly converge at the end of a campaign cycle, generally toward the Republican candidate.
When this occurs, Democrat media pollsters accomplish the shift in their topline results by shifting the demographic weighting of their respondent populations, then try to claim this represented a sudden last minute shift in the electorate itself. See, most infamously, the 2016 ABC/WP poll explaining why their double digit Clinton lead disappeared two weeks before the election when they started counting white working class voters.
Herding has fallen off the past couple cycles, though. The pollsters don't seem to care any longer whether their final results appear transparently partisan. Maybe they figure CA will manufacture enough votes after the election to make their final national topline results appear respectable.
Mr. W: "A smart layperson would assume these pollsters must be delivering something valuable though they don't know how they do it."
ReplyDeleteOnly Charlie Brown expects Lucy to hold the football for him to kick after she has pulled it away so many times in the past.
You would think even asses would at some point pick up on the fact they are being gamed by their political party.
"First, I DO know how they do it."
ReplyDeleteAn engineers knowledge of survey research, lol.
"The dirty little secret thing of value, is results that influence public opinion. "
Everything comes back to a secret conspiracy for these people. And this one is as silly as the others. For one thing it contradicts his talk of herding and he doesn't even get that.
Again, this is not a serious person.
You would think even asses would at some point pick up on the fact they are being gamed by their political party.
ReplyDelete# posted by Blogger Bart DePalma : 2:52 PM
These poll numbers are GREAT news for John McCain!! - Sniffles DePalma
"Herding has fallen off the past couple cycles, though. The pollsters don't seem to care any longer whether their final results appear transparently partisan"
ReplyDeleteLol, here's the tell! It's just results driven drivel for the partisan talking point of the moment.
"These poll numbers are GREAT news for John McCain!! - Sniffles DePalma"
ReplyDeleteI guarantee a Romney victory! Sniffles DePalma
BTW, as I noted in a past thread, the "umpire" metaphor in some sense goes back to the Founding. The issue really is that it undersells the discretion/power in fact applied.
ReplyDeleteMr. W:
ReplyDeleteI believe you post at 538.com under another alias.If so, fellow Democrats there have previously educated you about the existence of "herding," if not how pollsters accomplish the phenomenon.
the point I'm trying to get across here is that the theory of statistics really IS based on an assumption of high response rates.
ReplyDeleteAgain, for a large population such as 130 million voters, your are wrong.
"I believe you post at 538.com under another alias.If so, fellow Democrats there have previously educated you about the existence of "herding," if not how pollsters accomplish the phenomenon."
ReplyDeleteOur Bircher of course missed the point, which was about how he switches from arguing relying on the existence of herding to dismissing it at the point it he needs to to make his partisan point of the moment. Every square fact he deals with is simply hammered into the round peg of his partisan talking point of the moment. And this guy acts like he knows the intricacies of polling! Heck, he can't even catch on to this simple point.
This is not a serious person. This is a partisan incoherent.
Let's be clear: "Herding" is a theory. Sample sizes can be used to predict, probabilistically, an upper bound on sampling error. The inequalities used don't work the other way--not with anything like the same accuracy.
ReplyDeletePersonally, I love the way Bart pulls out pollsters that he thinks are more accurate, based on their results being what he likes. Oddly, they both get mediocre scores on 538's poll scoring. Apparently Bart has no understanding of the "stopped clock" phenomenon.
"Sample sizes can be used to predict, probabilistically, an upper bound on sampling error."
ReplyDeleteGreat Ceasar's ghost, you're from bizzaro world, aren't you? Where everything is backwards.
Sample sizes can be used to predict, probabilistically, how sampling errors should be distributed, given certain assumptions. The only upper bound on sampling error is 100%.
ReplyDeleteAnd, as any real pollster would tell sampling errors aren't the only source of error, or necessarily the largest. They're merely the only sort of error that can be straightforwardly calculated.
Brett, can you quote a theorem that produces a lower bound on the difference between a sample mean and the mean of the underlying population? Just wondering...
ReplyDeleteClearly, an inhabitant of Bizzarro (sic) world would think a person in normal reality was weird...
C2H5OH: Let's be clear: "Herding" is a theory
ReplyDeleteHerding is a repeatedly observed phenomenon. There are multiple theories as to its cause. My theory comes from comparing the demographic spreads of polls before a sudden shift in the topline results during the last couple weeks before an election. They tell the tale.
Personally, I love the way Bart pulls out pollsters that he thinks are more accurate, based on their results being what he likes.
I noted the only two polls which pegged the results of the 2016 election. FWIW, I also questioned Ras's R+1 2018 generic over at 538.com before the 2018 election. I suspected Ras assumed the 2018 electorate would look a lot like the 2016 electorate, when historically during midterm elections the out party has higher turnout than the ruling party.
Oddly, they both get mediocre scores on 538's poll scoring.
Oddly, 538.com's top ranked pollsters were some of the worst in 2016.
538.com only compares each pollster's final poll with the election results to derive it's pollster rankings, which ignores all the previous propaganda polling and assumes pollsters will "herd" toward reality at the end.
The only upper bound on error is 100%, and the only lower bound on error is 0%. That's part of what it means to be "probabilistic". The calculations tell you what the distribution of the errors should look like, given certain assumptions which real world polling seldom satisfies.
ReplyDeleteMr. W: Our Bircher of course missed the point, which was about how he switches from arguing relying on the existence of herding to dismissing it at the point it he needs to to make his partisan point of the moment.
ReplyDeleteI did not raise the topic of "herding" here, I merely corrected the others' misdefinitions of the phenomenon, then noted the phenomenon was receding over the past couple cycles.
Let's review here.
ReplyDeleteAs our Birchers really do, start at the end, which is also the beginning. Trump and his supporters are angry about recent polls showing him losing. They think these polls will dispirit GOP voters. Now, what Birchers do when they come across a fact that they think disadvantages them is: deny the fact exists. Now, often, the fact is put forward by a whole lot of experts with far more training and experience in the subject matter than they. So they have to argue that all those experts are wrong and they are right about this subject they know less about, in fact it's often asserted that the experts are involved in a conspiracy. What's the motivating conspiracy? Why the starting point, to hurt the GOP of course (remember, this is what made the fact necessary to be challenged for them in the first place), of course! Remember, they did this around the Romney campaign (surely Obama was more unpopular than the polls said). They did this in talking about climate change. They did this when talking about Covid. And they're doing it now. It's what they do. Wacky conspiracy theory is how they understand everything. Everything and one is as partisan as they are, professional expertise and standards don't exist, and they are under a constant barrage from almost every institution in the US. It's an amazing combination of paranoia, arrogance and festering victimhood.
Evidently, Brett has never heard or understood how "confidence intervals" and things like the "Chebyshev inequality" work. Sure, in probability world, "upper bound" merely means in a "probabilistic" sense (that is, with a certain probability.)
ReplyDeleteBrett thinks this means they don't work at all, which would be a terrible shock to any statistician or probability theory student...
And just as the arguments they make are structurally essentially the same, so are the holes in them. An inability to not only consider but imagine what should be obvious countervailing evidence and counter-examples, a hyper-skepticism toward the findings of people with far more training and experience in the fields they opine on, an incredibly naive set of assumptions about the coordination necessary for the conspiracy, etc.,.
ReplyDeleteSurvey research is a complicated field, but one that nearly every major organization and politician pays big money for because it produces valuable results. The people doing it know what they are doing for the most part, even when they are 'wrong' they are remarkably close to being accurate generally. They would have little interest in publicizing inaccurate results to 'influence public opinion' because their reputations and pocketbooks will suffer (note, many pollsters do both internal and public polls). Look at Bircher Brett's incredibly sloppy but run of the mill conspiratorial thinking about publicly released polls: well, polls you don't pay to see *must* be trying to influence you because why else would they be done? The simple, obvious answer of 'well, they're usually done by news agencies or universities whose raison d'etre is to discover and disseminate knowledge' cannot even be imagined by Bircher Brett. Or another example: if people are just paying for the polls to make them look good or their opponents look bad, why are polls so generally consistent between the many organizations doing them? Again, the standard conspiracy type thinking is deployed to deal with this otherwise very inconvenient fact: oh, well, they're all lying in order to seem like they're not so *plainly* lying! This kind of thing would be incredible if it weren't such standard conspiracy type thinking. What is interesting is this really is the GOP's base go-to style thinking on just about everything, as we've seen here over the years.
*The simple, obvious answer of 'well, they're usually done by news agencies or universities whose raison d'etre is to discover and disseminate knowledge'
ReplyDeleteBtw-I don't by this mean to suggest some naive idealism, news agencies make money by 'scooping' their competitors on discovered knowledge about current events/affairs and colleges exist off research monies and tuition dollars which are tied to their reputation for being able to put out accurate findings on a variety of subjects. This of course only strengthens my point, of course. Pollsters have massive incentives to be accurate and they generally are so. It's much more believable that the people who know what they're doing re polling are just right and Trump, who was one of the few Presidents elected without a majority of the popular vote and is beset with at least two massive crises right now is actually had a dip in his popularity (especially relative to Biden who has not yet been the subject of what will surely be a massive campaign to discredit by the GOP). But conspiracy theorists are going to conspiracy theory. It's what they do. These are not serious persons.
Mr. W: Trump and his supporters are angry about recent polls showing him losing. They think these polls will dispirit GOP voters.
ReplyDeleteDemocrat media propaganda polls inflating support for Dem candidates and policy positions have been part of the landscape for decades.
The only Republicans who take this propaganda seriously are establishment types who support progressive policy and are constantly in fear their own conservative/libertarian base's opposition to such policy will cause them to lose elections. the conservative/libertarian base has largely stopped playing the game altogether, which is part of the reason why pollsters miss so many of these "shy tories." I hang up on pollsters. Others I know lie to screw with them.
Trump's main attraction to his fans is his unmerciful trolling of the Democrat media. Propaganda polling is low hanging fruit in this trolling.
If the Dem media intends this polling to encourage Dems to vote, Republicans to stay home or both, I see no evidence they are succeeding. 2020 looks to be the most serious disconnect yet. Trump routinely draws overflow crowds of 40,000+ to rallies across the country, while Biden was lucky if he could fill a High School gymnasium before they hid him in a basement. Trump supporters are twice as likely as Biden supporters to tell pollsters they are enthusiastic to vote for their man. Finally, Trump is drawing far more primary votes than any previous unopposed POTUS seeking reelection. BUT, we are supposed to believe Democrat media polls claiming Biden enjoys a 9% average lead over Trump.
What constantly amazes me is how, after these polls repeatedly raised and then elections repeatedly dashed your partisan hopes, and in the face of the disconnect I just laid out that you still believe this propaganda.
"Democrat media propaganda polls inflating support for Dem candidates and policy positions have been part of the landscape for decades."
ReplyDeleteAll the best conspiracies are long running!
"The only Republicans who take this propaganda seriously are establishment types "
And seemingly Birchers who have to concoct conspiracy theories to try to discredit ones they think are dispiriting to their cause, like in this thread. Or also sitting Presidents who threaten legal action against those who do such polls...
"Trump's main attraction to his fans is his unmerciful trolling of the Democrat media."
I'd agree Trump is fundamentally a troll and his fans are sick people who enjoy such.
"I see no evidence they are succeeding. 2020 looks to be the most serious disconnect yet."
These poll numbers are great for McCain!
"Trump supporters are twice as likely as Biden supporters to tell pollsters they are enthusiastic to vote for their man."
I thought polls were propagandist bunk? Remember, I said Bircher Bart would tout polls when he liked the results. He did it in this very thread. Lol.
"BUT, we are supposed to believe Democrat media polls claiming Biden enjoys a 9% average lead over Trump."
It's certainly more plausible that a President who has never shown much popularity has had a relative decline as he presides over two major crises (Bircher Brett admitted the other day something to the effect of that he'd be surprised if Trump didn't lose what with all that was going on) against a candidate who has been out of the news recently than that there's a long running conspiracy of which this poll is the latest. Well, to a non-conspiracy prone nut that is, to them the latter is of course standard fare.
"after these polls repeatedly raised and then elections repeatedly dashed your partisan hopes"
Polls have consistently been very accurate actually, even the 'wrong' ones in 2016. Bircher Bart doesn't have a premise to stand on apart from the sloppy logic he uses with them.
Bart apparently thinks that people here have had their "partisan hopes" dashed by polls which failed to predict the outcome. Oddly, I only recall one election, 2016, where that could even be said to be true (and, speaking for myself, the disappointment was more because of events that occurred late in the election -- not because I was invested in the polls.)
ReplyDeleteOne wonders how Bart determines things, like "Trump supporters are twice as likely as Biden supporters to tell pollsters they are enthusiastic to vote for their man."
Through polling, perhaps? The only other possibility appears to be from the voices in his head...
This comment has been removed by the author.
ReplyDeleteMy question for Bart. If Trump loses, will you blame it on voter suppression by Democrats, achieved by talking bad about Trump? According to you in 2012, this was how Obama beat Romney.
ReplyDeleteC2H5OH: Bart apparently thinks that people here have had their "partisan hopes" dashed by polls which failed to predict the outcome. Oddly, I only recall one election, 2016, where that could even be said to be true
ReplyDeleteDuring the blogging age alone, I recall the Dem poll driven denial followed by howling and gnashing of teeth during 2000, 2002, 2004, 2010, 2014 and 2016. Like a broken clock, Dem media polling is correct when Dems have good years.
Larry Koenigsberg said...My question for Bart. If Trump loses, will you blame it on voter suppression by Democrats, achieved by talking bad about Trump? According to you in 2012, this was how Obama beat Romney.
ReplyDeleteIf Trump loses in 2020, it will almost certainly be because the economy. Absent a recession, voters have reelected every POTUS seeking a second gig.
Dems always talk trash about their opponents and vis versa. Team Obama was particularly effective in suppressing the swing state working class vote for Romney by micro targeting those voters below the radar with messaging portraying Romney as an evil plutocrat. They did the reverse with their own voters to get them out to vote. Extremely effective in shaping a very ahistoric electorate.
Along those lines, Team Trump has a terrific tech team putting the Obama social media GOTV and fundraising efforts on steroids. Trump fan boys and girls have to register for rally tickets online and provide their contact information. Over a million did so for this weekend's rally alone. About a fifth of them are not regular and/or are minority voters. This has provided Team Trump with millions of potentially new voters to bring into the electorate this cycle. The DNC is privately very worried about Trump's tech edge. I haven't read anything about whether Team Trump is employing this edge to suppress the Dem vote.
"The DNC is privately very worried about Trump's tech edge."
ReplyDelete-- and of course, Bart would know...
This comment has been removed by the author.
ReplyDeleteWe are 2 Appletinis away from Sniffles guaranteeing a Trump victory.
ReplyDeleteBTW, the liberals over at Faux News have Biden +12.
ReplyDeleteI think today's 'I know these Dem polls are inaccurate conspiracies because just look at the Dem poll results on enthusiasm' by Bircher Bart should go into the Bircher Bart Hall of Classic Goofs and Rake Steps.
ReplyDeletehttps://m.youtube.com/watch?v=2WZLJpMOxS4
Dilan -- of course, you mean "lie to all the other pollsters" -- because candidates rely on polling to adjust their messages and take the pulse of their campaigns, and you would probably be fired by your boss at the request of all her other advisers...
ReplyDeleteI actually don't think politicians need polls to do messaging. They can figure it out. And one reason I know this is because- my own profession figures it out. Yes, at the very high end of trial practice there are jury consultants and things, but most trial lawyers and all appellate lawyers manage to craft our messages without polls and persuade people.
Polls are a crutch at best, and really not even that, because the samples are so small and the policy ones usually don't tell you something you don't already know.
Presidential polls have very large populations, so the response rate does not matter .
ReplyDeleteI'm no polling scientist, but that should only be a true statement if the non-respondents share the same distribution of political views as the respondents.
If they don't- if, say conservatives, or liberals, are more likely to hang up on a pollster than the other side, then the pollster is hosed.
Political polling is, on average, consistently remarkable accurate (even in 2016 for President they were very close to being on target).
ReplyDeleteThat depends on how you define "accurate".
In fact, there's plenty of evidence of major inaccuracies in 2016. What is that evidence? Very simple- all the up-and-down fluctuations over the course of the election campaign. There's no way there were all those dramatic fluctuations in what was actually pretty clearly a very stable race.
I don't think we should look at a graph with 100's of inaccurate individual polls and say "see, the final one was pretty accurate!". One of the big problems is that the tracking polls don't really track anything except noise.
Roberts doesn't like to be treated like a rubber stamp. He'll approve anything the admin wants if it's done the right way but he doesn't take to being told he has to believe obvious, stupid lies.
ReplyDeleteI don't think any judge likes being lied to. Some judges will put up with it more (and to be clear, if the law is absolutely clear that the lying side should win, the lying side will still, usually win, sometimes with an admonishment), but "not saying stuff that your judge thinks is BS" is an excellent litigation strategy whether you are seeking a liberal or a conservative result.
Survey research is a complicated field, but one that nearly every major organization and politician pays big money for because it produces valuable results.
ReplyDeleteMaybe. Or maybe they pay because there is very little you can control in a political campaign so the things you can control, you pay big money for.
The same way a big deep pockets litigator will pay big money for a Supreme Court boutique firm when cert is granted, even though it doesn't really affect the probability of success very much.
I think the basic notions of persuading the public are pretty well known, and do not actually require a lot of survey work (much of which is scientifically questionable) anyway. James Carville didn't need a poll to figure out "the economy, stupid", even though he surely had polls in the campaign.
But campaigns have lots of money to burn, so you might as well spend it on polling. Demand drives supply.
BD: "The DNC is privately very worried about Trump's tech edge."
ReplyDeleteC2H5OH: "-- and of course, Bart would know..."
This is hardly a state secret. Do some reading.
Start with Thomas B. Edsall, “Trump Is Winning the Online War" (NY Times) for background.
Then google "Brad Parscale," Trump's social media wunderkind.
Mr. W: I think today's 'I know these Dem polls are inaccurate conspiracies because just look at the Dem poll results on enthusiasm' by Bircher Bart should go into the Bircher Bart Hall of Classic Goofs and Rake Steps.
ReplyDeleteYou really don't have a clue, do you?
Dem media poll over-sampling of Dem leaning demographics and, thus, under-sampling GOP leaning demographics changes the topline horse race results, but has no effect at all on questions about subjects like enthusiasm within those demographics. If the poll reweighted demographics to reduce the number of Trump supporters from 300 to 250, the same percentage of those supporters would remain very enthusiastic about voting for their man.
Dem media poll over-sampling of Dem leaning demographics and, thus, under-sampling GOP leaning demographics changes the topline horse race results
ReplyDeleteAre we back to the "skewed polls" narrative that explained why the 2012 polls ere wrong and Romney was going to win?
Dem media poll over-sampling of Dem leaning demographics
ReplyDelete# posted by Blogger Bart DePalma : 11:26 PM
Faux News has Biden +12 you dimwit.
there's plenty of evidence of major inaccuracies in 2016 [...] all the up-and-down fluctuations over the course of the election campaign. There's no way there were all those dramatic fluctuations in what was actually pretty clearly a very stable race. [...] One of the big problems is that the tracking polls don't really track anything except noise
ReplyDeleteIf the tracking polls were only tracking noise, the fluctuations wouldn't be dramatic. If on the other hand, the fluctuations are dramatic, we have no idea whether the tracking polls were accurate or not prior to election day.
if, say conservatives, or liberals, are more likely to hang up on a pollster than the other side, then the pollster is hosed
Of course that is true, but there is no evidence the polls are biased. Thus, it is likely the polls do a fine job with a 6% or 16% response rate.
Then google "Brad Parscale," Trump's social media wunderkind.
ReplyDelete# posted by Blogger Bart DePalma : 11:11 PM
Or you could Google Joseph Goebbels. That's probably your best source for Trumpism.
ReplyDeletehttps://maps.google.com/url?q=https://bloginbits.com/
https://www.google.fr/url?q=https://bloginbits.com/
https://www.google.es/url?q=https://bloginbits.com/
maps.google.nl/url?q=https://bloginbits.com/
maps.google.pl/url?q=https://bloginbits.com/
maps.google.com.au/url?q=https://bloginbits.com/
images.google.co.in/url?q=https://bloginbits.com/
https://maps.google.co.in/url?q=https://bloginbits.com/
"I'm no polling scientist, but that should only be a true statement if the non-respondents share the same distribution of political views as the respondents."
ReplyDeleteI keep pointing that out, and they keep claiming it proves I'm innumerate. LOL!
It's only believable as a matter of faith that the non-respondents do not share the same views as the respondents. Why shouldn't they? Because he wishes that they wouldn't -- and that their "true feelings" fall in with what he thinks they should. How could such a claim be verified? Through polling in which the non-respondents were forced to give honest answers? It is to laugh.
ReplyDeleteBrett, of course, has said that the upper bound on sampling error is 100 percent -- which, of course, displays his innumeracy. That could only be true if the sample size were 0.
It is typical of people afflicted with selective perception that they are incapable of dealing with probabilistic situations, because they see the occasional incidents as typical. This is also why they will pull out particular polls and particular pollsters as worthy of mention, while denigrating the vast majority.
All this discussion has strayed from the actual issue: the fact that Trump is a very unpopular president who, at least at present, is growing less popular week by week, and that this affects his ability to sway decisions and actions. Note that it's not merely a large segment of the public who think little of Trump; a rather large segment of the people who have worked for him also seem to have negative feelings for him.
"I actually don't think politicians need polls to do messaging."
ReplyDelete"In fact, there's plenty of evidence of major inaccuracies in 2016. "
"But campaigns have lots of money to burn, so you might as well spend it on polling. Demand drives supply."
Dilan's comments really boil down to a kind of standard 'I don't really know what I'm talking about here, but here's a possible Slate-level hot take turning the conventional wisdom and practice on its head!' (whenever you see someone end with taking a truism and turning it on its head you should be wary that this is what's going on).
I mean, again, it's not just politicians who use polling. Marketers, advertisers, the US military, most major corporations use it both in house (surveying employees) and for public polling. The government uses it for things like consumer confidence. It's silly to think it's providing nothing of value, that all these disparate organizations in disparate situations are just burning that money. If it were so, and if a 'non polling scientist' lawyer like Dilan could figure this out by simply musing, then organizations in fields where polling is common would start to not buy into 'the lie', and save their money (also, if polling is actually inaccurate they'd make *more informed* decisions without them), this would give them an edge in these very competitive markets and poll-reliant firms would lose out. But of course this hasn't happened. There's been an *increasing* reliance and use on polling across the board. It's people going with just 'their intuition' and this supposedly 'obvious' messaging that are harder and harder to find.
And, again, in general, it's actually remarkable how accurate the polls are. Dilan tries to point to the fact that polls fluctuated during the 2016 race as some kind of proof they are not. This assumes that the end result was going on the entire time. If what is meant by that is that historically races tend to tighten as 'leaners' 'come home' to their party in a two party system, well, sure, most races tighten up (as noted 'blow outs' like Carter are still within 10 points). But there's actually every reason to think there will be fluctuating polls for a race like 2016: Trump had trouble winning his own primary and was polarizing even within that electorate. Heck, Bircher Bart here called him a fascist about up until he was elected POTUS! It makes sense that his numbers had to climb as GOPers and leaners warmed to him (or became more horrified at a Hillary victory). In the end, in the perfect 'reality test' we have of poll performance, they were remarkably accurate even in this year that is held up as the epitome of polling failure.
So, the very premise upon which the entire conversation is built is a false one. And the reasoning it's built upon-either conspiracy theory or Slate-hot-take reasoning, is just not good. These arm-chair quarterbacks will of course feel like they know something important that everyone else involved, those with much more at stake and much more training/education/experience, will continue to opine that everyone with those things is engaged in a mass delusion or conspiracy. A sensible person asks themselves what's more probable, that the latter are all nuts/conspiring or the armchair quarterbacks don't know what they're talking about.
"but has no effect at all on questions about subjects like enthusiasm within those demographics"
ReplyDeleteBircher Bart of course forgets castigating these 'Dem polls' as useless and ignored by GOPers, that they hang up on them, lie to them, etc., hopelessly skewing the results. This would be as true for enthusiasm results.
Again, Bircher Bart knows the Dem polls are useless because look at the results of the Dem polls! I said he'd do this at the start, and he sure did. Just like he argued for herding and then dismissed it. Because the only principle he was ever operating on is: what supports the GOP in this current discussion?
"I keep pointing that out, and they keep claiming it proves I'm innumerate. LOL!"
ReplyDeleteThis is typical Bircher level thinking. 'I know a little something about a basic theory involved in polling work, and there's reason to think that some current aspects of the polling landscape are problematic for the theory, so I can conclude that the polling work is fatally flawed.' You see it with them all the time (climate change, economics, etc.,). I mean, sure, the basic theory is complicated when you have politically relevant groups opting out or that are missed in polling attempts at higher rates than others (this is the lesson of the famous Methods 101 discussion of the Readers Digest Landon-FDR telephone poll). But what's incredible is for an armchair quarterback like Bircher Brett to assume that the people with far more experience and training in the relevant field *don't know about this* and haven't spent a lot of time and thinking about how to address this. I mean, when people went from landlines to cell phones every survey researcher I know remembers the huge discussion about what to do with the fact that young people were increasingly less likely to have land lines and they were increasingly less likely to answer phone surveys than older people. But survey research didn't stop. It continued, in fact it flourished, and all manners of organizations increasingly used it. I mean, maybe consider that this is a complex field and they figured out ways of dealing with the current wrinkle the way they've figured out many other wrinkles along the way that you just don't get or understand because it's the inner workings of a field you don't know much about (considering, as I've said, the polls are remarkably accurate and that people increasingly rely on them, this seems like a pretty good bet, huh?).
If the tracking polls were only tracking noise, the fluctuations wouldn't be dramatic. If on the other hand, the fluctuations are dramatic, we have no idea whether the tracking polls were accurate or not prior to election day.
ReplyDeleteThat assumes the only way you can tell if public opinion is changing is a poll.
However, that is not true. We have other metrics. For one thing, reporters talk to voters. There are lawn signs. There's activity and enthusiasm on social media. There's the news cycle and how many people are reading or watching.
None of those things showed that people were over and over again, flip-flopping and changing their minds between Hillary Clinton and Donald Trump. It was ONLY the small sample, statistically invalid "tracking polls" that were showing the constant fluctuation. They were noise.
Of course that is true, but there is no evidence the polls are biased. Thus, it is likely the polls do a fine job with a 6% or 16% response rate.
You are using "there is no evidence" to then claim the null hypothesis is that the polls are accurate. That's not true. The null hypothesis is that we don't know anything about the electorate. The polls are advancing a hypothesis and therefore have the burden of proof. That's why we require a 95 percent confidence interval!
So part of polling has to be to prove that the non-responders are comparable to the responders. A pollster doing actual science could do this by using other mechanisms such as personally interviewing people who don't respond to telephone surveying, and publishing the results in a peer reviewed journal. Until you know that, however, the poll is basically junk science.
What the polling industry is actually based on is demand. People don't want to wait for science to actually establish these things are accurately measuring public opinion.
Good comments MW. Also thanks to C2H5OH and others for the focus on the math.
ReplyDeleteIt's only believable as a matter of faith that the non-respondents do not share the same views as the respondents. Why shouldn't they?
ReplyDeleteThis is complete, utter BS. The matter of faith is the opposite.
WHY would you expect, in this polarized world, that the non-responders would be ideologically the same as those who do respond. Heck, there's an obvious reason right now why they might not be, in that there's a conspiratorial movement out there which doesn't trust pollsters. But beyond that, there are other things that might skew it in the other direction too, such as the fact that caring about your privacy IS ITSELF A POLITICAL BELIEF that should correlate with other political beliefs.
You WANT polls to be accurate, because politics is your hobby and it is fun to throw around statistics, just like a baseball fan does. But that doesn't mean this is sound science, any more than WAR statistics can calculate the influence a team member has in the locker room at getting his teammates to cohere as a unit.
As I say above, pollsters have the burden of proving the ideological mix of non-responders is the same as responders. And that can actually be determined by taking several months, using other methods to find out about non-responders, and then publishing that information along with the polls. The problem is that people want the poll results instantaneously, before the science has been done. Which makes the whole thing... basically unscientific.
But beyond the issue of burden, it is absolutely the pollsters who are acting on faith here. CRITICS of polls are just saying "you don't know anything about the non-responders", which is a true statement. YOU are saying "why wouldn't they be the same?", taking it on faith. This sort of burden shifting is just irresponsible.
BD: Dem media poll over-sampling of Dem leaning demographics and, thus, under-sampling GOP leaning demographics changes the topline horse race results
ReplyDeletejust_looking said...Are we back to the "skewed polls" narrative that explained why the 2012 polls ere wrong and Romney was going to win?
Actually, the 2012 race illustrates a number of my polling points.
Use the Real Clear Politics polling chart and follow along.
Registered Voters: : Polls of anyone self identifying as a RV reweighted by census polled demographics. The primary error here is the demographics of the population at large when compared with the actual electorate overcount the young and minorities - Dem leaning groups. As 538.com recently admitted, these polls overcount Dems by “a few points.”
Observe the spring and summer RV polling projecting enormous Obama leads far in excess of the actual results.
Likely Voters: This is where the pollster guesses what they think or want the demographics of the electorate to look like, then they reweight actual responses by these guesses. In 2016, the Dem media pollsters wrongly assumed Obama’s young and minority base would turn out for a white establishment Dem and the white working class would stay home as in 2012. They were wrong.
Observe the post-labor day LV polling again projecting Obama leads up to twice the final results.
Herding refers to when Democrat media pollster results suddenly converge at the end of a campaign cycle, generally toward the Republican candidate. When this occurs, Democrat media pollsters accomplish the shift in their topline results by shifting the demographic weighting of their respondent populations, then try to claim this represented a sudden last minute shift in the electorate itself.
At the beginning of October 2012, the Dem media polls suddenly collapsed their previous wide Obama leads into statistical tie by adding in a historical percentage of white working class voters they were previously excluding.
What the Dem media pollsters and nearly everyone else outside of Team Obama did not know was the Dem targeting would cause roughly 6 million white working class voters who tossed out the Dems in 2010 not to vote for any POTUS candidate in 2012. In the face of this ahistoric electorate, the Dem media pollster herding actually created a small systematic error in favor of the Republican Romney.
Dilan's comments really boil down to a kind of standard 'I don't really know what I'm talking about here, but here's a possible Slate-level hot take turning the conventional wisdom and practice on its head!'
ReplyDeleteNo, I used an example of something I do know about (the marginal value of Supreme Court boutiques) to show that when tons of money is available, people will spend it on stuff even if the marginal utility of it is not very high.
Much of what polls tell us is stuff I could have told you without polls. Some of what polls tell you is inaccurate- AS EVEN POLLSTERS CONCEDE. So the question is how much is really in the middle, the sweet spot: stuff that is both accurate and novel. You are saying it is a lot, but it isn't a slatepitch to suggest that even if it were not a lot, campaigns would still pay lots of money for it.
Marketers, advertisers, the US military, most major corporations use it both in house (surveying employees) and for public polling.
There's a big difference between most market research and political polling, which is time and frequency. United Airlines doesn't call a random sample people up every 3 days asking their opinion of the brand. The "tracking poll" is a phenomenon of politics, not market research, because politicians are playing in this time-limited game where they use polling as sort of the scoreboard. That is not the case in corporate America.
And that has huge implications for the validity of surveys. First, the refusal rates are higher, because the public is absolutely sick of pollsters. Indeed, the fact that up to 95 percent of the public hangs up points out something else, which is that polling people is actually a huge intrusion and very rude. Most of the public hates political pollsters. That, right there, is a reason to stop taking polls. Personally, I would have more moral character than to go into a business where 95 percent of the time I am bothering people who don't want to be bothered. I'd stop. Taking money for polling is basically blood money.
But beyond the moral point, the refusal rates mean that you aren't as sure about this sort of data as you are about market research. And further, corporations aren't trying to measure the day to day fluctuations. Which means political polling also has smaller samples. Many pollsters will survey less than 500 people in a tracking poll. That's just terrible! And it's nothing like corporate market research.
And, again, in general, it's actually remarkable how accurate the polls are. Dilan tries to point to the fact that polls fluctuated during the 2016 race as some kind of proof they are not. This assumes that the end result was going on the entire time.
As I point out above, the notion that all these voters were flip-flopping over and over again in the 2016 campaign is something that you would have to be a total idiot to believe. Seriously, the media was talking to voters all that time. We had social media. The flip-flopping was not occurring, because it didn't show up ANYWHERE but in the tracking polls.
Again, this is your hobby. You like politics. It's a sport for you. You can't stand the idea that the scoreboard may be inaccurate. So you bring yourself to believe obviously untrue things about politics so you can have your little scoreboard. It's ridicuolous.
Having said all that, where Bart and Brett go wrong is assuming the polls are biased against their position.
ReplyDeleteIt's perfectly clear that political polling is psuedo-science and its proponents massively overclaim their accuracy. It's also perfectly clear that it is an immoral industry that constantly invades the privacy of the public, which just wants them to go away.
But it's not clear that there's a specific bias in one direction or another. Indeed, most of what I am describing should manifest itself simply as noise and meaningless fluctuation, along with occasional whoppers (such as the pollsters' utter failure to project Sanders' win in Michigan in 2016 or Trump's wins in the Rust Belt in the general election).
To conclude polls are biased against the Republicans, you have to engage in the same psuedo-scientific hocus-pocus that the proponents of polls engage in, making hasty conclusions based on small samples and circular reasoning.
"None of those things showed that people were over and over again, flip-flopping and changing their minds between Hillary Clinton and Donald Trump. "
ReplyDeleteCan you cite scientific evidence from the sources you mention-yard signs, reporter interviews, social media (talk about methodology challenges!) that the race was as stable as you claim? Or how about some evidence that such measures produced a reliable prediction of an outcome, say a study of those things right before the election that came as close as the polls did?
BTW- it's important to note what 'fluctuating' early polls are showing. If Hillary is up by 10 points when Donald has just won the nomination that's not saying right leaners are choosing Hillary over Donald. It usually means the undecideds are large. And that's to be expected. As I said, you had Bircher Bart here saying Donald was a fascist at this point in the election. He certainly wasn't going to vote for Hillary, he was going to answer undecided or other, but very likely come back home with enough pro-Trump stuff in conservative media and anti-Hillary stuff. Likewise, when Hillary dipped it's likely that has to do with people who just were like 'Donald Trump? That goofball, I guess Hillary' but who were never very pro-hillary, and then when things like Comey's letter comes out they are like 'hmm, I always heard there was something fishy about her.' These people might be won back over by the end (that Donald is just too goofy as I thought).
Again, here's what we know: when there is a reality test for polls they tend to be remarkably accurate. If there's any evidence that other methods do as good a job as that I certainly don't know about that. People who have stakes in this almost always spend gobs of money on it. And people like Dilan or Bircher Brett don't know much about the field.
So, now it's time for inductive logic. Is it more likely that the stakeholders are crazy/stupid and the market miraculously doesn't punish them, that the trained experts haven't thought of ways to compensate for the issues arm-chair quarterbacks like Dilan or Bircher Bart see as fatal, and while the polls are remarkably accurate in general when we have the ultimate reality test, or more likely that Dilan and our Bircher's are wrong, engaging in conspiracy theorizing or Slate-hot-takism? I think I know what I'd put my money on.
And one parting shot. It's the arrogance of people who like polls as their hobby that pisses me off the most. People get into it because using statistics makes you feel scientific and superior to everyone else. The thing is, bad science gives you the same feeling of arrogant superiority as good science does.
ReplyDeleteIndeed, talk to any serious global warming DENIER and you will hear the same arrogance. That they are in the know because they've looked at this study or that hockey stick or whatever, and they can PROVE their position. They have NUMBERS.
I am NOT claiming that all political polling is inaccurate, or that it's all biased. But it doesn't meet the criteria of actual science. Science requires care, effort, the willingness to say you don't know when you can't get a good data set or enough good information to test a hypothesis. Most of all, science requires TIME. It requires that you have enough time to actually do the work, with care, and eliminate all the alternative hypotheses.
Political polling is about getting the numbers out as fast as possible, because that's what sells. That's what sells to the campaigns, and that's what sells to the media, and that's what sells to you, the hobbyists, ultimately. It's a business. And like any business, it is looking for suckers.
Instead of congratulating yourself for having such a "scientific" approach to politics, ask yourself why you are so willing to set down what should be the skepticism of an actual intelligent person in favor of buying this product. Ask yourself why you want this product so bad.
Because that's what the polling industry really is. It's a product designed to sell well to the type of people who like things that look scientific. And it turns out, there are a lot of marks out there.
C2H5OH: It's only believable as a matter of faith that the non-respondents do not share the same views as the respondents. Why shouldn't they?
ReplyDeleteThe better question is: Why would they?
People are not interchangeable ciphers. People differ by personality, gender, culture, economics, faith, ideology and, apparently, by their willingness to participate in polls.
The phenomenon of polls undercounting groups who tend to vote "conservative" has been observed over too many decades, in too many countries by too many pollsters to be dismissed as a fantasy.
"I used an example of something I do know about (the marginal value of Supreme Court boutiques)"
ReplyDeleteWhat do you know of Supreme Court level litigation? Have you ever argued there? You might want to take the advice you give above re: law professors in Con law opining on how administrative law works...
"You are saying it is a lot, but it isn't a slatepitch to suggest that even if it were not a lot, campaigns would still pay lots of money for it."
Why? Why wouldn't they spend that money on the things you say is better at indicating where the public is at? Studies counting lawn signs or reporter interviews or some such nonsense? They're just all en masse engaging in a mass delirium and mutual disarmament?
"There's a big difference between most market research and political polling, which is time and frequency."
As someone who has done internal polling work for corporations I can tell you that your comment "I'm no polling scientist" is surely correct based on what you say in this section. If anything, response rates are *lower* when you interrupt someone's dinner to ask them about what kind of air conditioner they think more highly of. With politics there is broader interest and a civic commitment to participate.
"Seriously, the media was talking to voters all that time. We had social media."
Talk about pseudo-science!
Can you cite scientific evidence from the sources you mention-yard signs, reporter interviews, social media (talk about methodology challenges!) that the race was as stable as you claim? Or how about some evidence that such measures produced a reliable prediction of an outcome, say a study of those things right before the election that came as close as the polls did?
ReplyDeleteI don't need to. I am defending the null hypothesis!
But no, I am not going to do your research for you, especially since this thread closes in a couple of hours and I have to go back to a brief I am writing. However, I assume you watch media coverage of presidential campaigns like everyone else does, and when public opinion shifts, it is fairly easy to see in media coverage. (For an example, see the last 3 weeks on policing issues. It's reflected in the polls, but you could have figured it out without polls.)
BTW- it's important to note what 'fluctuating' early polls are showing. If Hillary is up by 10 points when Donald has just won the nomination that's not saying right leaners are choosing Hillary over Donald. It usually means the undecideds are large.
Except the undecideds were not large in 2016. The tracking pollsters claimed they were small, which would have meant that people were actually fluctuating. They, after all, ask voters if they are undecided.
Now, if voters are lying and saying they are undecided when they aren't, well, then YOU have a problem, because that means why should we believe anything the voters are saying?
Again, here's what we know: when there is a reality test for polls they tend to be remarkably accurate.
1. Only if you only consider the final polls. 2. "Remarkably accurate" ignores that they get tons of things wrong, including not showing Trump's success in the Rust Belt in 2016.
Is it more likely that the stakeholders are crazy/stupid and the market miraculously doesn't punish them, that the trained experts haven't thought of ways to compensate for the issues arm-chair quarterbacks like Dilan or Bircher Bart see as fatal, and while the polls are remarkably accurate in general when we have the ultimate reality test, or more likely that Dilan and our Bircher's are wrong, engaging in conspiracy theorizing or Slate-hot-takism?
1. What you are calling hot-takism is just standard intelligent person's skepticism, which is DEMANDED when something is claimed to be scientific. Remember what the null hypothesis is in science. The polls have the burden of proof. 2. I think polls are a product, and should be evaluated like a product rather than as science. If you understand polls to be a product, you understand that it's possible to sell things that provide little actual marginal benefit. Happens all the time.
What do you know of Supreme Court level litigation? Have you ever argued there?
ReplyDeleteI've filed or worked on probably 35 Supreme Court cases over 20 years of practice.
At any rate, the Supreme Court boutique issue has been written about by numerous lawyers and law professors, so it's not like I'm the only one who says this.
Why? Why wouldn't they spend that money on the things you say is better at indicating where the public is at?
They do. Campaigns have BILLIONS of dollars. They can spend on everything. That's what makes it like Exxon going to the Supreme Court!
As someone who has done internal polling work for corporations I can tell you that your comment "I'm no polling scientist" is surely correct based on what you say in this section. If anything, response rates are *lower* when you interrupt someone's dinner to ask them about what kind of air conditioner they think more highly of. With politics there is broader interest and a civic commitment to participate.
That's only one subset of what I said. I said that your corporations have time to construct a good survey. They also use other techniques to find out about what the public thinks, including focus groups, personal interviews, customer feedback forms, etc. Polling is a small part of a much larger picture.
Talk about pseudo-science!
I agree social media is psuedo-science! It all is. The problem with you is you think political polling ISN'T.
ith politics there is broader interest and a civic commitment to participate.
ReplyDeleteBy the way, this is a whopper. If up to 95 percent of the people are not responding, that's the opposite of a "civic commitment". That's "leave us the bleep alone".
Which pollsters should do. Their business model is less important than voters' privacy.
Of course some individual polls might be bad, just as some medical studies are. And polls working under time constraints will often be less valuable than ones that are not, just as medical studies under time constraints will be less good than longer term ones. None of this justifies your Slate-hot-take that polls are blood money industry and not scientific any more than the corresponding failings of medical studies mean that field is one of blood money and not scientific. Polls, like any scientific endeavor, are not perfect, not easy and sometimes wrong. But, in general they are remarkably accurate and give important information that only a fool and/or arrogant person ignores or explains away as conspiracy theory or mass delusion. That's why every serious person and organization across a multitude of fields pays good money for them regularly.
ReplyDeleteAs a small tip: if your pet theory requires you to believe that a mass of people spend a lot of effort as a result of a mass delusion or conspiracy theory and that you know some secret that people much more knowledgeable in the field don't, you should really check and re-check that theory. As a matter of inductive logic it's just far, far more probable that it is you making an error, biased, being uninformed about something key, etc.
Dilan:
ReplyDeletePollsters are faced with a nearly impossible problem because the cannot obtain truly random samples of the general population or the electorate because of mass non-participation and outright lying. They are forced to reweight their incomplete and distorted samples to resemble targeted populations.
If they stopped there, we would see a fairly large margin of error clustered around reality. However, we rarely see this.
There is almost always a systematic error on one side of reality. This means the pollsters are reweighting their samples in the same way to achieve the same desired populations and results.
This bias is hardly limited to political polling. We see it constantly in "climate change" and now COVID "model" projections.
Lying through statistics.
"Campaigns have BILLIONS of dollars. They can spend on everything. "
ReplyDeleteSo do their opponents, and if they spend their money foolishly and the other side does not, they increase their chances of losing. If polls were such bunk, over time campaigns would start to realize this, gain a competitive advantage, and polls would'nt be so widely used.
"They also use other techniques to find out about what the public thinks, including focus groups, personal interviews, customer feedback forms, etc. Polling is a small part of a much larger picture."
So do political pollsters. Of course polls don't provide the only useful information. But they do provide useful information.
"corporations have time to construct a good survey."
Again, just admit you don't know what you're talking about. Markets and situations are fluid and move incredibly fast. Corporations are constantly tracking these trends.
"I don't need to. I am defending the null hypothesis!"
No, you're claiming polls are wrong because their results don't track the methods you mention which you claim finds that there were not the fluctations the polls found. But of course you've 1. no evidence those other methods show that and 2. that those other methods are more reliable than polls
"Only if you only consider the final polls"
Which are the only ones that are ultimately 'reality tested.' And they're remarkably accurate. And they are conducted using the same methods as the earlier polls.
"it is fairly easy to see in media coverage"
This is silly, media coverage is much more likely to be ginned up with no scientific basis.
"What you are calling hot-takism is just standard intelligent person's skepticism, which is DEMANDED when something is claimed to be scientific"
ReplyDeleteThe epitome of hot-takism is when you have someone who is not trained in a complex field making an arm-chair conclusion that everyone in the field, and all the people who widely rely on it all the time, are getting it wrong. It's a classic fail of inductive logic: what's more likely, that all those experts are wrong/conspiring and those paying good money for it are stupid and markets aren't punishing them, or that the arm-chair quarterback missed something about a complex field they know little of?
Here the hot take consists in: I have a methods 101 idea about response rates and the logic of how they work in generalizations from samples. I know there are response rate challenges for polling work. Therefore I conclude the polling work is fatally flawed.
ReplyDeleteIf it were this obvious there wouldn't be a gazillion experts out there competing with each other to provide more valuable results than the next one to stakeholders who have everything at stake in knowing and pay good money for it. I mean, maybe they know about this challenge, have long thought about it and developed fixes that, as we see when ultimately tested, produce generally solid and remarkably accurate results.
Heck, polling work seems much more accurate than meteorlogy even to the layperson, but I bet even people like Dilan check the weather in the morning before going out (not to mention people with much more at stake, fishermen, people holding outside events, etc).
None of those things showed that people were over and over again, flip-flopping and changing their minds between Hillary Clinton and Donald Trump
ReplyDeleteThe polls said Clinton started with a narrow lead which grew to its maximum in mid-October (after the debates), and then narrowed back to its starting point after the Comey letter. That pattern doesn't strike me as flip-flopping over and over again.
You are using "there is no evidence" to then claim the null hypothesis is that the polls are accurate.
The polls are reasonably accurate as measured by their performance in predicting who won and by how much.
Dilan, so you would delay the release of a poll until the pollster determines, through whatever methods apply, that the non-respondents and respondents do not differ? That would, of course, defeat the purpose and utility of a poll, which is to take a "snapshot" of the electorate -- and, given that your theory that people's opinions do not change, and sometimes change rapidly -- make the results worthless by the time they are released.
ReplyDeleteIt's perfectly fine to suspect that a given poll, especially from pollsters who are not known for careful selection of their samples, is not accurate. And you may, if you wish, disbelieve entirely in the utility of polling.
Let's distinguish between the mathematical theory and the practice. You cannot argue with the theory. As to practice, the only methods that make sense to try and correct for systematic error are methods which rely on prior information, such as historical biases in differences between respondents and non-respondents. Naturally, if the differences in these populations changes rapidly, the pollster is going to be caught -- but if we accept your theory of a relatively unchanging population, that won't happen, at least often.
You can't have it both ways: either the respondents and non-respondents are pretty stable, or not. If so, then one can correct using history. If not, then why would they change one way on one poll one week, swing the other way a week later, and so on? Meaning in that case that there is no significant difference between the respondents and non-respondents, and the need to correct for it is pointless.
just_looking said...The polls said Clinton started with a narrow lead which grew to its maximum in mid-October (after the debates), and then narrowed back to its starting point after the Comey letter.
ReplyDeleteThe Dem media polls (without Ras, which always had a tight race, and later IBD) had Clinton leading by over 7 points, until herding to about 4 points at the end. The reason for this miss was all the polls assumed the young and minorities who cast ballots at historic levels for Obama would do the same for an old white establishment candidate (although none had done so in 2010 and 2014), while the white working class would continue to stay home as they did in 2012.
The Dem media poll herding was entirely driven by reweighting their demographics to something closer to historical reality, not Comey's letter concerning then months old story of Clinton's Espionage Act crimes . For example, closely read the ABC/WP poll's explanation for the complete disappearance of their double digit Clinton lead two weeks before the election, when they claimed the composition of the electorate changed overnight.
Where the Dem media polls admitted they really missed the mark were in the swing states, where they were all projecting a Clinton sweep. Remember all the nonsense about a "blue electoral college wall?"
The only cover the Dem media polls could find for their systemic debacle was CA manufacturing about 3 million votes in the weeks after the election, making their national Clinton lead projection look semi-respectable.
So according to Bart, pollsters deliberately degrade their product in order to ...?
ReplyDeleteAfter all, these people get paid based almost entirely on the accuracy of their results. (Some exceptions like the one Trump recently hired in order to give him results he likes -- which is likely to result in a short life as a pollster.)
Or, they're consistently incompetent -- which is an obvious case of psychological projection on his part. These people may not be the most intelligent people in the world, but they're far beyond his level.
Polls, to the true believers, are like Communism is to Communists. They can't fail, only we can.
ReplyDeletejust standard intelligent person's skepticism
ReplyDeleteThat's a fine standard, if done with humility and nuance.
I recall this being said too:
If we listened to the right people, and not the loud generalists, folks would have a better notion of what goes on (and this is true for other legal specialties too).
We have a lot of "loud" generalists (comment-wide) around even if each have specialties. I'm not sure who the "right" people are though on this thread a few seem more on the ball on this poll thing than others. One or more who I believe are the sort of "right people" experts an appeal to authority would lend an ear to anyhow.
"As a small tip: if your pet theory requires you to believe that a mass of people spend a lot of effort as a result of a mass delusion or conspiracy theory and that you know some secret that people much more knowledgeable in the field don't, you should really check and re-check that theory."
ReplyDeleteThat's not my theory. My theory is polling is a business and we should expect pollsters to act like anyone else who sells something. No conspiracy involved.
"The polls said Clinton started with a narrow lead which grew to its maximum in mid-October (after the debates), and then narrowed back to its starting point after the Comey letter."
ReplyDeleteThat's not true. The Comey letter fluctuation was the same size as several other fluctuations in 2016.
"You can't have it both ways: either the respondents and non-respondents are pretty stable, or not."
ReplyDeletePublic opinion was stable in 2016. It shifts sometimes, but there's no reason to believe polls are a particularly accurate mechanism for capturing those shifts.
Dilan, I think I'll take polls over yard signs and rally attendance, thank you. While not perfect, my guiding principle is: take what data you can, with salt as needed.
ReplyDeleteHaving said that, I'll add that I share your feelings about polling in general, and I never respond to polls.
"We have a lot of "loud" generalists (comment-wide) around even if each have specialties. I'm not sure who the "right" people are though on this thread a few seem more on the ball on this poll thing than others. One or more who I believe are the sort of "right people" experts an appeal to authority would lend an ear to anyhow"
ReplyDeleteThe relevant experts on polling would be political scientists who publish peer reviewed studies on actual public opinion (i.e., not the ones who study tracking polls).
I have only spoken to one such person in my life. He did not agree with everything I am saying here- he is probably somewhere in the middle of this discussion, but he definitely believed that most polls discussed in the media were garbage. (He said there are a few pollsters who use much better methodologies and publish much less often, and those are the polls he tends to consult.)
One thing is the political science types on Twitter tend to be pundits like Jonathan Bernstein rather than the folks doing a lot of peer reviewed publications. I would really like to hear from the latter.
The Comey letter fluctuation was the same size as several other fluctuations in 2016
ReplyDeleteCitation?
Do you agree that the polls have been reasonably accurate in predicting who wins and by how much?