Yves here. Because I find it painful to listen to pundits and soi-disant public intellectuals, I’ve managed to miss a lot of the revisionist history by the centrists and Vichy left to pretend that they were outside the hive mind confident of a blue wave and a Biden blowout. This post presents some examples. I hope readers will supply more in comments.
By Barkley Rosser, Professor of Economics at James Madison University in Harrisonburg, Virginia. Originally published at EconoSpeak
As I write this the outcome of the presidential election remains uncertain, although the trends seem to be heading steadily in Biden’s favor and it seems very unlikely he will lose, but the hanging on of this is beginning to get to many of us. Most other races have been decided, although some of those are still up in the air. In any case, although it is near certain Biden will win, it will be by a substantially lower margin that forecast by the main polls, with him not taking several states he was forecast to take. This is also mirrored in Congressional (and some state legislative) races, with it unlikely, although not completely out of the realm of possibility, that Dems will control the Senate as had been forecast as a 2 to 1 likely outcome by 538, and instead of gaining seats, the Dems lost some in the House while retaining control, with the exact number of losses also still a bit uncertain. In short, Trump and the Republicans did much better than generally forecast, and this has lots of people upset and lashing out at each other, including, reportedly, Dem members of the House.
No, I am not going to join the mob howling at the pollsters, nor am I going to join the mob howling at Schumer and Pelosi. I shall look at how I think the pollsters got it wrong, although I think they have been doing their best and trying to avoid certain errors they made in 2016, when they also underpredicted how well Trump would do. Nor am I going to join those dumping on Schumer and Pelosi. I think the Congressional outcomes have a big fat zero to do with anything either of them have done or said or not done or not said. It was all about the down-ballot GOPs gaining from the underpredicted surge of Trump supporters coming out to vote for him, the phenomenon that needs to be understood.
Who I am going to pick on, although not too much really, are some reporters and commentators I shall simply label as pundits, “whining pundits” to be precise. To name three who all appeared in the Style section of the Washington Post, not the editorial page, I note Monica Hesse, Hank Stuever, and Margaret Sullivan, roughly in order of how annoying and hypocritical I find their whining.
I note of these I have long respected Margaret Sullivan a lot, and she was mostly fairly factual, covering in fact the difficult problem of the pollsters messing up (“Who who won, but it wasn’t the pollsters”). In fact while accurately noting details of how off the pollsters were, she, who regularly covers the media, extended the critique to the broader media beyond the pollsters. She accurately noted the sparcity of coverage of the degree to which Trump was appealing to Latino males, especially those in South Florida. But she may have overdone it when she appeared to blame the media for relentlessly covering Trump’s mishandling of the pandemic, which she suggested allowed Trump to appeal to his followers by pointing to a hostile mainstream media. What she did not do when she came up with that was to note that she herself has much of the time in recent years dumped on the media for not being relentless enough in exposing the various flaws of Trump.
From his byline of “Critics Notebook,” Hank Stuever (“Difficult to tune out, but even harder to watch”) focused on the election night TV coverage. Much of this is reasonable, but he too turned to whining, not about the polls or even the coverage, but the naive optimism of sucker liberals among the public. He did this by quoting TV reporters who made sneery such remarks as the results began to turn unexpectedly so much for Trump. So we get “Long before the post-midnight madness, CNN anchor Jake Tapper reminded viewers that there has been a lot of magical thinking lately on the part of Democrats and liberal pundits, who spent too much time entertaining pie-in-the-sky dreams of tipping Texas and Georgia [which now appears likely to in fact end up gong for Biden] to blue states, imagining a landslide victory for Biden. “As they say, ‘You can’t get high on your own supply,'” Tapper zinged.” OK, but I am unaware of either Tapper or Stuever expressing such views prior to the election. Ooops!
Somehow Monica Hesse (“Crushed by a landslide that never heppened”) makes these other two look completely reasonable with her sneering whining, also lacking any previous reporting on how wildly over-optimistic all these people were. She accurately notes that election officials had warned of “an election week,” but then we had “in some liberal circles there was a jacked-up fantasy that nobody would have to wait for Pennsylvania’s tortured ballot count because by 10 p.m. Joe Biden would turn Florida and North Carolina blue [this latter remains a possibility, although not too likely] – and maybe Texas, why not?” As an added dig she adds that “for the past years, the demographics in my inbox who most fervently believed in a 2020 blue landslide were White liberal men and occasionally White liberal women. Surely, they insisted, what had happened in 2016 was a blip.” This may be right, and she does then argue that Black women and others did not have such illusions.
OK, for all the hypocritical whining and sneering by these folks who did not issue these warnings previously, the problem seems to boil down to observers taking polls seriously. 538 had a national gap of 8% nationally for Biden while Real Clear Politics had it at 7%. A likely outcome will have Biden at 4-5% ahead. This is a larger gap than in 2016, when the final 538 predicted national advantage for Clinton was 3%, with her getting a 2.1% lead. How then did things go so wrong in 2016? Ah, the errors varied across states, with some having the polls on the money, but they messed up badly in the crucial three states of Wisconsin, Michigan, and Pennsylvania, especially the first, and they were what provided the surprising outcome of a Trump electoral college victory.
The national gap this time looks to be somewhat larger, but the variation across states also seems to have reappeared. Some have been on the money, such as the respected Selzer poll in Iowa that nailed the Trump lead almost on the head. But then we have Wisconsin again, where one poll in the last few weeks actually had Biden ahead by 17%, with the overall averages having Biden ahead by 5-6%, but in the end having less than 1 percent lead (again, thank you Prairie du Chien!).
For myself on Facebook in comments I speculated that due to a likely higher turnout with such intensity on both sides, there was a larger variance to the likely outcome, with the 538 projections of the expected for Biden too optimistic, but with indeed there being also a non-trivial chance of him getting a landslide. And we should note that if 538’s main forecast had happened, 8% lead, this would have been viewed as a landslide, with indeed both Florida and North Carolina for Biden by 2% and with Texas a tossup, despite the whiny sneering by Monica Hesse. It was not just obviously pie-in-the-sky fantasy to take such an outcome seriously, although arguing it as an inevitability clearly was so.
As it was, even though I thought the pollsters had made some appropriate changes since 2016, with one for a bias about educational levels widely reported, I worried about “silent Trump supporters,” and I think those in the end were out there, especially among the better educated, despite the awfully loud and proud shouting by many Trump supporters. So I generally lopped 2% off the forecasted projections, but that was still not enough by a percent or two, although it does actually lead to pretty close to accurate the outcomes of which states would go which way.
I shall note that besides election officials one group that was also cautious about this and preparing for a high turnout by Trump’s supporters has been those in the Biden campaign. Many criticized him for not going to Texas and some other out-there possible states (although he made a few such trips). Of course, he laid low a lot in his basement, possibly too much. But when he was out he very heavily concentrated on the core three states that unexpectedly flipped in 2016, and it looks like this was a good decision, with Wisconsin and Michigan flipped back, and hopefully and likely Pennsylvania as well.
64 comments
Comments are closed.
One factor that I think pollsters frequently overlook is that effective campaigns don’t focus on populations as a whole – they know full well that there are only small chunks of an electorate who can be made change their mind and who can have an effect on the results, and so they focus their local ground game resources on these groups. These cohorts may be so small statistically that it doesn’t show up on broader polls if they’ve been successfully ‘shifted’ by one party/candidate or another – it only shows up in the final result. It seems to me that the Republicans are particularly effective at this, perhaps because they have a better ground game.
It may be that pollsters didn’t ‘miss’ South Florida hispanic males, so much as the Trump campaign quietly focused on them and successfully ‘turned’ enough to make a difference.
Good take. Pollsters think of themselves as actuaries when they’re really more like bookies. Except the mob would have gotten all new bookies after the 2016 debacle.
Polling is not seeing the forest for the trees. I’ve gotten several polling calls and was dropped once I answered initial questions about party, sex, age, education, ethnicity, region (and more). The margin of error becomes higher when selecting who to poll.
+1. (As in it happened to me too, one evening when I decided to see what a caller whose ID suggested it might be a pollster wanted to hear from a Green Party candidate. They never quite asked that specific question, but at some point they apparently decided they didn’t want to know who I was going to vote for….)
Trump campaign’s door-to-door canvassing [a chapter from Sanders’ book] in states such as Texas and Florida was not ‘overnight’; the MSM, to my knowledge, was either unaware of it or remained ignorant of its implications – just like the Democrats. For the latter case, see
https://www.commondreams.org/news/2020/11/05/omar-and-tlaib-credited-major-factors-securing-biden-victories-minnesota-and
I doubt if Omar and Thaib would receive due credit for their critical contribution to Biden’s taking Minnesota and Michigan.
Anglos love to think of Hispanics as one monolithic block. This glosses over the fact that the term Hispanic covers a lot of ground. Here in Texas, it mainly means people of Mexican origin although many have been born here and/or lived here all their lives. They often see things very differently from Cubans and Venezuelans in S. Florida.
I had the pleasure of taking a Latin American Culture class, a required elective I needed, and it was excellent. It was taught by a Prof with a Mecican-American background and I was only 1 of 2 Anglos in the class. The rest were from a Cuban, Puerto Rican, Mexican, Haitian, Dominican, and various Central and South American backgrounds.
One of the prime focuses of the class was the fact that they were anything but a “single block”. They had totally different attitudes about each other’s ethnicity, democracy, capitalism, socialism, and just about everything else under the sun… the single common thread was importance of family relationships.
Every one of these so-called pundits needs to be the sole Anglo in a class like this before they dare open their mouths again.
It’s not that important to guess an election outcome in advance, is it? Especially in a place like the US, where the range of possible outcomes is known with high certainty (the greens won’t win this year), but the outcome within that range is always uncertain.
Unlike many questions in life, this one will get answered just by waiting. And in the mean time, what can you do with an improved estimate? Whether you put the estimate of a Biden win at 90% or at 30%, is there anything in your life that you can do different? You still have to prepare for either outcome.
Of course, if you have a better estimate than others, you might make some money from gambling. But that is a zero-sum game, w can’t all improve our lives from gambling right. And polling or punditry is not about that – after all they don’t keep their insights secret, they scream them from the rooftop.
Professional campaigners have some use, but that is a small group of people. Even for politically active laymen, the best path forward is to ignore polls and try your best for your candidate.
So what s the point? Where comes the desire to guess outcomes come from?
Zamfir
November 7, 2020 at 5:03 am
I agree – well said and well reasoned. If you don’t think polls are accurate, disregard them.
I do think people get angry with polls because they believe they have an agenda and are purposefully wrong.
I believe the real reason is more maudlin – the MSM now is essentially a 24/7 operation, and they just have to fill all that time. FACTS, at least the ones that MSM managers think the public will gaze at, aren’t all that many (well, a zillion facts – the media could reproduce all the laws, rules, and regulation regarding postal voting, but who would watch that for say, even 1 minute?) And all the facts one could be interested in are available on the internet, generally from the original source.
Speculation about the future, however, is limitless.
So the MSM is pretty much just entertainment. I treat it the same way I treat spots – occasionally entertaining but generally inconsequential.
Polls serve a lot of purposes: shaping expectations, exciting the base, raising money, knowing where to spend money, A/B testing positions, putting pressure on disapproved candidates to drop out, fostering a simpleminded horse-race discussion that journalists prefer, etc. Mostly stuff that only benefits campaign managers. But for many it just satisfies a need to feel control over uncertainty.
I fully agree that it would be better not to have polls at all or not until the very end. I’d rather the election be the only poll. But there’s to much money involved to leave it until then.
Yes, I agree with this. I feel my outsized ire at polls this time around was the clear attempt at yet again trying for opinion-making. Now that I’m thinking about it, it’s not even that, it’s angling (not so subtly) for opinion influencing while claiming to be a bloodless techno-activity.
Uh, it sells ads. These things aren’t put out there for the cognoscenti.
I did see one modeller involved in the Economist forecast suggest that if they had suggested Biden got Hilary plus 2%, in 48 states it would have done better than their sophisticated model. For what it’s worth, that blog had a lot of hand-wringing in advance. Their model, and 538’s, suggested high 90s% of Biden winning but they didn’t seem to believe it was that strong. And, even though Biden appears to have won, they are saying they didn’t do very well. The challenge is that although there actually only a small number of elections they can really draw on statistically and many have “special circumstances”. Which suggests that the models may be indicative, but we probably will never be able to place a lot of confidence in them.
The takeaway is that something worked for the Democrats in Arizona and Georgia.This was true up and down the ticket in those states.
That magic ingredient in Arizona and Georgia was on the ground community organizing by LUCHA and others in Arizona, and by Fair Fight Action and others in Georgia.
So TV ads were largely a waste but grass roots community organizing worked.
Sample size of one, but in NH which went for Biden, D senate and house but R for govenor, state legislature, congress and executive council: I have a friend who ran D for a state legislature seat in a district that has been R since the dawn of time. He lost his first race, narrowly lost a special election when his opponent resigned, won in 2018, and won this year. He went door to door in every town in his district. He talks to people about issues, and a non trivial portion of his votes came from people who have never voted for a democrat in their lives.
> So TV ads were largely a waste but grass roots community organizing worked.
It was pointed out yesterday that as much was spent on an individual D Senate race (McGrath v McConnell, IIRC) as on the entire nation’s D state legislature races, and the Ds lost badly in both.
At some point, the pundits may figure out that the Ds (at least the ones who are in a position to control the direction of the Party) don’t actually want to govern.
There’s this belief, I guess, that elections can be bought by the expenditure of ever more money to addle the proles into going for one or another candidate or position. Seems to be true in some instances, but as with McGrath, silk purse does not obscure sow’s ear even with an opponent apparently as unpopular at home as McConnell. Would seem better to get the issues right, and have a candidate who is at least not disgraceful or obnoxious and who can be seen, rightly or wrongly, as bound to stand behind positions taken on issues that matter to ordinary citizens.
Interesting that the turnout appears to have been pretty high all across the country. What activated people to get out and vote? Even though, of course, the “choice” was largely only between shades of sh!t between slices of Wonder Bread ™.
Slim checking in from Tucson.
In addition to the organizing that Ray just mentioned, here’s another factor that I think was at play:
Arizona Republicans overplayed their hand. They brought us politicians like former Maricopa County Sheriff Joe Arpaio, former governor Jan Brewer, and our current governor, Doug Ducey. Not to mention that much-reviled SB 1070, which got a lot of that LUCHA organizing underway.
And then there’s the PMC factor. I’ve mentioned it here before.
In this year’s election season, Tucson’s PMC neighborhoods were rocking the Biden/Harris signs like nobody’s business. I’ve never seen anything like it before.
In past elections, our betters have gone for presidential candidates like, well, Donald Trump. And Mitt Romney. And John McCain. And George W. Bush.
Community organizing isn’t just a technique, it’s a form of thought, an ideology. Something motivated the people who did the work in Arizona and Georgia. I don’t know what it was, but it probably wasn’t the Democratic Party’s donor class. I imagine, if you’re going to overcome endemic tribalism and superstition you need something pretty obvious and forthright, a set of positive material goods like Medicare for All, and not conspiracy theories, fables, culture war, and hatred. Fat chance, I suppose.
Suggest search-and-replace all instances of “donor class” with “Briber class” or “Owner class.” The money in politics is never an eleemosynary gift. Strings always attached.
The magic ingredient in Arizona was Trump. By disrespecting their war hero senator, John McCain, he helped give the state to the democrats.
irony alert: Democrats did better in AZ GA beacuse they got out of their Covid-bunkers and knocked on doors (in part due to local lockdowns being lifted).
Precisely the opposite of the Biden campaign at the national level.
The “polling problem” is one I hate to hark on about but is crucial and has only once been seriously addressed by a major polling organisation – in 2017 by YouGov in their “alternative” model of the 2017 UK General Election. This, interestingly, was the ONLY mainstream model to predict correctly that PM May would lose her majority. At the time I applauded them, but warned that though they correctly used a “secondary dataset” to “solve” the statistical problem, their choice of method in collecting this data would at some point break down horribly. As it happened it then broke quite spectacularly in the very next election (2019) and so I fear the baby will be thrown out with the bathwater when YouGov were actually on the right lines (which I said – strongly – in 2017).
Here’s the issue. People (indeed I saw in a comment on NC 2 days ago) think that aggregating polls reduces margins of error. This should be listed in those Buzzfeed lists of “Facts” Everybody Knows That Are Actually Total Lies. The fact such aggregation has still NOT correctly got USA predictions right in a bunch of general elections recently “must” show nefarious activities – WRONG. Now, I don’t think there were nefarious activities in this election – at least of the type the Trumpites claim. The “nefarious activities” are more that the polling organisations REFUSE to poll using correct methods as known since 1985 because of their funding masters, laziness and fear of striking out from the pack. “Specification Error in Probit Models”. Adonis Yatchew and Zvi Griliches. The Review of Economics and Statistics, Vol. 67, No. 1 (Feb., 1985), pp. 134-139. Is the crucial theoretical statistical first paper proving the crucial property which is the following in lay language: when you aggregate DISCRETE data (like polling data – “Democrat or Republican”) then failure to do some very specific data testing and if necessary adjustment first, will NOT “iron out” errors in estimates and “reduce the margin of error”. In fact aggregation – as done with a CONTINUOUS outcome (like blood pressure to improve point estimates and reduce error) – does NOTHING to reduce error and will frequently EXACERBATE it. For the statistical gurus, heteroscedasticity on the latent scale – in this case a person’s “internal voting scale with definitely Democrat at one end and definitely Republican at the other” leads to BIAS in point estimates IN AN UNKNOWN DIRECTION AND OF AN UNKNOWN MAGNITUDE. It is because if we observe 7 out of 10 voters in a sample voting Democrat, it can arise because of a mean effect (everyone knows exactly what they want, won’t “jump around” etc) or a variance effect (a mild pro-Democrat preference happened to translate to a much stronger observed effect due to large variance) or something in between. And if you don’t know which of these “explanations” is true, adjust for it, and ONLY THEN aggregate, then your aggregation means an effective sample size of 100,000 might give you results at least as wrong as some rubbish poll of 1,000 people; the “biases” don’t cancel out. TL;DR – if you don’t correct heteroscedasticity in polling data first then “more data” doesn’t produce better results, often the opposite.
This problem was further expounded across mathematical psychology, branches of applied microeconomics and other fields for 15 years with perhaps the best exposition given in Journal of Econometrics 89 (1999) 197-221. “Combining sources of preference data”. David Hensher, Jordan Louviere, Joffre Swait. This really shows how Stats programs always assume a “7 out of 10” is a MEAN effect. If it’s a variance effect – and it’s not unreasonable to assume variances are large in things like this US General Election where Biden is concerned – then your predictions may be horribly wrong and way too “pro-Democrat” in reality.
One last point that anyone who bothers to read the small print at the bottom of the probit/logit sections of the manuals for Stata/SAS etc will/should know which alerts you to all these problems: mathematically an observed frequency is, in the likelihood function, concealing a PERFECT confound between MEAN and VARIANCE. ALL the stats programs “solve” this problem – equivalent to an equation like xy=8 – by setting the variance (or a function of it) to be one. Thus they simply go, OK y= 1, so x=8. This is DREADFUL but we do it every time we run a logit or probit model. Really you only “solve” the problem with a secondary dataset (like x=2) – that’s what YouGov did (albeit badly in terms of generalisability over time).
Although I dislike the Economics “so-called Nobels” – McFadden did deserve his because he knew what the problem was and that he needed a “second equation” (obtained from real observed preference data and trade-offs) to “solve for x and/or y” and thus predict demand for BART in SF before it was built. Nobody in polling gives a hoot. And thus here we are…..Groundhog data every 4 years.
Excellent comment; thanks for that write up.
You’re welcome and thanks for reading and engaging. I sometimes think that it’s understandable that nothing is done about polling. Then I think “what the family blog?” This is a fundamental weakness! Then I remind myself it took until 1985 to publish the paper showing the likelihood function has the mean (beta terms) and variance being tied together as a multiplicative term so maybe things really do move slowly in this field.
Indeed, we’re taught “rules” in stats classes like “heteroscedasticity causes inconsistency but point estimates remain unbiased”… Not realising this is NOT true for discrete outcomes – it causes BIAS! Stats teaching is lacking.
Yes thank you!
Again, thanks! I’ve not always been good at explaining or being patient (particularly when suffering from poor mental health in the past) but now I have left the field ironically it can be easier to explain the issues.
Part of me wishes the 3 or 4 others who “could have” helped explain these issues had done so 5 years ago. I think you need 4ish “biggish names” to push through a paradigm shift (like MMT always seemed to have). But they for various reasons weren’t on for it. Plus, though on a personal level I’d love to change things, I think other issues NC either espouses or has a lot of time for like MMT and land economics reform are more important so I can’t complain.
Thank you very much for this Terry, a fascinating contribution (although I confess that this depth of statistical analysis is way beyond my skillset). I’ve always had a suspicion that many aggregation techniques can amplify rather than smooth out errors, I’m glad to see that there is some confirmation of this.
Thanks and each time I engage I get more ideas how to explain things. Imagine getting 10 blood pressure readings from an individual when you know they’re normally distributed. You could fairly easily “construct” the normal distribution with the right level of spread and peak to fit these. Thus you know mean and variance.
Now imagine 10 votes from the person (here we must imagine 10 universes or some such). We don’t OBSERVE the “political scale” directly anymore. We see 0s or 1s. We use a statistical “link function” (logistic for logit, normal for probit) to INFER the curve. But we only know 70% were 1s. We can do this with either a (normal if probit) distribution that is “very peaked” but not too close to cutoff point (so 70% is above the cutoff point)…. OR a distribution that’s is “very flat” but centred at DIFFERENT point (but still with 70% above the cut off). If it flattens then the centre must move “up” the scale to maintain 70% above the cutoff point.
How do we know which is correct “explanation” for 70% Democrat vote? WE DON’T. Discrete data provides us with one equation with 2 unknowns! But pollsters assume everyone is only differing by MEANS not variances. As soon as you and I have different variance we can’t be aggregated til the variance differences are accounted for. Pollsters don’t do this. This matters because information/events affects means in some cases…. And in other cases variances…which makes the situation even worse.
OK thanks, I finally get it. Thanks for sticking with the explanations.
maybe the polls were correct and the vapor voting machines were, er, wrong.
Terry Flynn
November 7, 2020 at 7:11 am
very elucidating. Reminds me very much of a quote attributed to Keynes:
“worldly wisdom teaches that it is better for reputation to fail conventionally than to succeed unconventionally.”
It also reminds me of an essay in which Feynman details all the ways rats in corridor experiments were found to be incorrectly performed – but after finding out, the people who run rat run experiments (pun intended ;) just don’t bother to perform the experiments correctly
http://calteches.library.caltech.edu/51/2/CargoCult.htm
Thanks! Yes my (now former) field is largely called “discrete choice experiments”. The stats was “relatively easy”. Learning how to devise human experiments was HARD! Humans are sneaky and very capable of spotting things. It’s also very much an art (given that you often don’t have the key “2nd dataset” to unconfound mean and variance. So you must do lots of research, qualitative work and liaison with experts in the particular field you are investigating).
TF – might I ask you to clarify this idea? It sounds to me as though the problem is that when people are polled, they are, to an extent, guessing how they will vote. It might then be that people who said they would vote B were less accurate than people who said they would vote T and that someone with your knowledge would seek to test and correct for this. Have I got the right end of the stick, please?
Exactly right. The B people are less certain but pollsters don’t register this. Net result is that “more of the distribution” goes to the T side. On voting day the T vote is higher than expected. I’m a Brit but I learnt a lot about attitudes to B from here….. Many (non NC) people, I suspect, knew little about him hence the uncertainty.
Thanks. Plenty of food for thought about the human condition then. This site continues to be an education.
Thank you! This is enlightening.
I’m not sure that it is a good idea (scratch that; it’s a downright bad idea) to give the duopoly parties better measurement tools, but perhaps there is a future in “underground” consulting to verifiable progressives. How to fund that is another question.
Again, thank you!
I don’t have any sympathy for the pollsters and as others above have said, I don’t even see what purpose they serve for voters as opposed to the media and political parties who clearly benefit from them.
But that said, I think opinion pollsters done a fair job of the essentially meaningless task of revealing public opinions towards the candidates, if you consider your opinion to be how you represent yourself to other people you don’t know very well. For example a ‘shy Trump’ voter, by defintion, doesn’t want you to think they’re in favour of Trump so is it really correct to label them as a ‘Trump supporter’? I don’t think it is although it’s essentially a philisophical question. My guess is that voting choices has become so controversial that there’s an increasing divergence between people’s ‘opinions’ and how they’ll vote, rendering the whole polling process increasingly pointless
For example a ‘shy Trump’ voter, by defintion, doesn’t want you to think they’re in favour of Trump so is it really correct to label them as a ‘Trump supporter’?
Yes as you are what you’ll do when put in a position of power to act on it. Most shy voters aren’t that way in spite of having power, they are that way because they are in insecure relationships where the cost of revealing their true beliefs is higher than cost of staying silent.
The true opposite of shy Trump supporter aren’t the boisterous Trump supporters, it’s the Biden voters that lay out every single argument for crossing the line but never do even when they are in the position to do so. They don’t do that because they are afraid to cross the line for social reasons, they do that because they care far more about impressing the intellectual elites than they do waging political power and that would be true even if they held the highest office in the land.
Captain Renault: “I’m shocked, shocked, that the polls wildly overestimated a Biden blowout.”
Everyone had their priors before the election. It seems now nearly everyone has the same priors, but harder and louder.
I believe it was Lambert who said we should differentiate between “shy Trump” and “lying Trump” voters. I know this 20-20 hindsight, but it seems like it would have been easy to identify 1000 (or some number) Trump voters through yard signs or social media postings and poll them. I know, the family member who put out the sign might not be the person who put out the sign, but doing the same thing with an equal number of Biden supporters would have given some clue to the percentage of both lying and shy Trump voters. Unless we really believe it was strictly a Trump thing rather than a conservative “f*** the liberal media” thing, it seems like it would be a good idea for 2024.
How does it benefit a Trump voter to give the snobs and scolds a straight answer? After four years of hearing about how even their dogs are racist why would Trump voters bother with anything other than the culturally correct answer?
I didn’t say they had any ethical obligation to tell the truth. I was aware that “lying” is an emotionally charged word, but it’s also the correct one, and I’m pretty sure that I was using the term that I saw used on water cooler earlier this week.
The main thing is, I don’t think it’s just a Trump phenomenon. If a measurable percentage of conservatives want to f*** with the liberal media by lying to pollsters rather than just refusing to respond, this would be away to get an idea of how badly it’s skewing the polls.
What Terry Flynn said above is probably much more important, btw.
Thanks for the shout out. Incidentally your other points are valid too and ones I spent 20 years learning how to minimise (though not in a voting context). I dealt a lot with people who are members of online panels. They get points (redeemable into real goods) for doing surveys. This instantly sets up malign incentives (click through, lying etc).
I was personally involved with constructing lots of tests (see my google scholar publications) to identify this. To cut a long story short, you must switch the format of questions rapidly, measure time to click etc. Whilst it is easy for a human to “look like a Democrat” if they’re asked simple support/don’t support questions, it’s much much harder to do so if you have to “match a correlation matrix” based on choices between competing options. “Lying Trumpers” can easily be spotted. Because we find it hard to think in multiple dimensions.
I personally regularly sent the panel company the IDs of respondents who clearly were gaming the system. Their points were voided.
And just so nobody thinks I was over eager to void points. People with low levels of education routinely display choice probabilities (frequencies) close to 1/n where n is the number of options on offer. I never penalised such behaviour even though it added little to a study. To do so would be unethical IMHO. But maybe that’s why I became unpopular….. I wasn’t a git. Only people clearly trying it on and were like 10 standard deviations from normal behaviour with no plausible reason for doing so other than “messing me about” got into trouble.
The few times I’ve actually been contacted by a pollster I’ve hung up before finishing because the number of questions combined with dissatisfaction with the choices made it too aggravating. I suppose the length of the polls is partly a result of the need for control questions, but that probably has its own problems. Anything that limits the amount of people who are willing to finish a poll shouldn’t be assumed to be neutral. Of course, the the failure of pollsters to do the correcting that you mentioned, makes me wonder if that’s the reason for poll length.
If it’s to get a more picture, that makes the choices problems even more aggravating. For example: Obamacare polls that didn’t allow to disapprove because it didn’t include single payer, or the NYT poll that the writer of the Independent article in today’s links misunderstood*. It asked readers whether they thought racial inequality was a major issue without asking about increasing wealth and income inequality in general.
*I’m pretty sure he got this wrong:
. One poll suggested 91 per cent of Biden voters thought racial inequality was the most important issue, against only 8 per cent of Trump voters.
https://www.nytimes.com/interactive/2020/11/03/us/elections/exit-polls-president.html
In the town of 11000 registered voters where I live in Massachusetts, 3200 (28%) voted for Trump a bit more than in 2016. Lawns and fences throughout the town were covered with Biden/Harris posters. I walk and ride through the town almost everyday and saw only 2 Trump/Pence signs.
I voted for Trump in 2016 because Hillary’s Russian rhetoric scared the ever-loving crap out of me. However, this was purely symbolic (and I knew this at the time) for I live in the deep blue sea known as California.
In 2020… I had a really hard time deciding on who to vote for on the top of the ticket. It was NEVER going to be Biden because I still cannot imagine a bigger F-U to left-leaning voters than Biden and his “top cop” VP who is both likely to take over the Presidency due to Biden’s… deficiencies and who dropped out of the primaries for performing so very badly. All of this points to the Democrats coming out of the shadows as a right-wing party based on these two individuals track records alone.
Once again, in the deep blue sea, my top-of-ticket vote didn’t matter. So what did I do?
Well… as Mark Hamel said 2 weeks from the election “I’m counting the days until this election is over so I can try to ignore politics & go back to tweeting show-biz trivia, toddlers laughing & heartwarming animal-rescue videos like this one. #TwoWeeksToGo” https://twitter.com/HamillHimself/status/1318678397782949889
This represents the worst possible outcome to me; the left goes back to sleep and power is once again unchecked/unquestioned in Washington. Trump is horrible (namely because he continues to do most of what America did before, just without any verbal filter) but we’re all watching. So… while it made me sick to support either of the clown-car candidates given to us to “choose”
from (in the same way the North Korean’s are given a “choice” to vote for the Kim regime), I threw my vote away in Trump’s direction in the hope that if he were to win he’d have one more person giving him some measure of legitimacy.
I think this… F the Machine… protest vote was quite common.
Plus 100!!! Thank you – you said it better than I ever could.
That Mark Hamill Tweet was the most perfect “saying the quiet part out loud” moment I saw this election cycle. I think BLM and a lot of other groups are going to wonder where their white allies went about six months from now.
I think something important is being missed here: the relationship between the hucksters who run the Democratic party and the pollsters. I strongly believe the pollsters — like big accounting firms — do as they’re told and routinely juice their polls (usually with cleverly worded questions that force responders into yes-no-undecided responses that do not at all reflect their real views. I’m also reminded of the Monmouth poll director who while in Iowa trash talked the Sanders campaign for not participating in a Klobucharesque cheerleading contest in Clear Lake.
Increasingly, I’m seeing Trump as Jesse Ventura II. The people are speaking but no one is listening. We’re going to have to elect a face-painted Juggalo before the media acknowledges how strongly Americans are rejecting the leadership class. Or maybe someone who’s legally changed their middle name to [unprintable in a family blog].
But mostly I wonder why anyone watches any kind of televised “news.” Or reads a daily newspaper (w/the exception of the rare good articles which you’ll then see all kinds of links to because there really isn’t much good stuff out there worth sharing)?
They watch the news and read the paper since there isn’t anything better.
It’s a common error to think that in capitalism all products have to be good to find buyers: you just have to be better than a some others. And if you have more money to spend on glitz it’s easy to make it appear better even when it’s the same product.
Actually, I the only reason I listen to MSM news to see what propaganda they’re feeding us.
It’s 12 noon here on the East Coast. The AP just announced Biden as winner. The Pundits will continue to walk back on many things, but beam that they have pretty much convinced America that Biden is their man. As a Lefty, I should be happy. I am filled with gloom at the upcoming war machine’s gearing up to continue Global Dominance for the benefit of the few.
maybe somewhere between gloomy and happy…
should be amusing to hear what plans they have been storing in the dry powder room as they start to trickle out
Interesting, definitely!
Biden is the only Dem. who clearly won.
House: lost seats
Senate: To be determined, but unequivocally far from 2008
Statehouses: Lost seats
And all of this with a pandemic AND millions of long-term unemployed AND record turnout.
Nancy failed. She was too busy trying to protect the donors instead of her TRUE Constituents. She should have acted on Trump’s proposal, and added a little more. Trump would have pushed the Repubs because he would have been flattered it was, “His” plan. Instead, she played Ostrich so that nobody got a damn thing during a PANDEMIC!!!! All those Mom and Pop shops, people out of work, families having to home school with no money to buy teaching aids, extra food for all day at home, and the list goes on. SHE BLEW IT!
“Nancy failed” is too charitable, since it assumes that what you consider to be “success” was regarded to be desirable by her.
It looks like a political calculation — deny the Pres’ a “win” that would make many people grateful. And this might in fact be the difference that put JB over the top.
I would say, echoing others who have previously noted this, that on this legislative matter, NP “succeeded” in terms of what was important to her.
With the job that LUCHA did in organizing on the ground in AZ and GA it behooves the democratic party to move those people in its leadership up the food chain in the DNC to key strategic positions to initiate a 50 state strategy to win congressional majorities. However, this assumes the democratic party leadership has half a brain.
DC Dems are just as classist DC GOPers, most of those field operation folks won’t moved to the big leagues cuz their diplomas are from the wrong schools or because all the nepotism candidates did not get a job yet
YMMV
Time to test them in Georgia – NOW.
Sanders lingo doesn’t work. Neither does self-righteous woke nonsense(with Bernie slowly adopted to his weakness).
Its sorta like not understanding the rustbelt came in 2 waves. The first was the post-WWI consolidation as automation surged and what took 15 plants only took 1 outside a big city. This killed little manufacturing county/towns over the midwest during the Great Depression and it never came back by 1950. The 2nd wave was the debt/consumption wave featuring offshoring in larger towns. If you don’t get that, you don’t get the problem. Texas white vote is slowly moving more Democratic again because the kids are leaving the small rural areas for work, changing voting habits. If you aren’t growing, you aren’t voting.
Building growth takes work on the ground, not baseless slogans like “no more fracking” “free college” or “Black Lives Matter”. I think one of the biggest moves this election was the continued fraying of the northern black vote and Democrats. Its not stopping either. Nor should it. They aren’t really into the party.