The Off Topic Topic

[Valmy]

What is fairly interesting from Brazil is it is another case of a polling miss in a major election. Sort of like in the last US election, the expected blowout didn't happen but the favorite still won (or likely will win), but it is something else to shake confidence in polling.

They predicted Lula would get the most votes and he did. Was there ever a point in history where polls were consistently nailing every election down exactly? Because I don't recall this.

[Josquius]

Poor Brazil.

Lula is not an improvement over Bolsonaro.

          

Your eyes are going to fall right out of your head if you keep doing that.

I'm no expert on Brazilian politics, but I do follow them more closely than many other countries due to my sister-in-law and my nieces.

Replacing Bolsonaro with a 76 year old who was convicted of corruption is not an improvement.  My saying so is not an endorsement of Bolsonaro.

Convicted of corruption by a corrupt judge allied with bolsarnaro who has now been given a plum position in government.

[Barrister]

Convicted of corruption by a corrupt judge allied with bolsarnaro who has now been given a plum position in government.

The prosecution of Lula was almost certainly politically motivated.

But I've heard no serious claims that the accusations against him weren't actually true.

[alfred russel]

What is fairly interesting from Brazil is it is another case of a polling miss in a major election. Sort of like in the last US election, the expected blowout didn't happen but the favorite still won (or likely will win), but it is something else to shake confidence in polling.

They predicted Lula would get the most votes and he did. Was there ever a point in history where polls were consistently nailing every election down exactly? Because I don't recall this.

Holy shit valmy, they were wrong and I can't even fathom why you are upset about this....

All the polls after the last debate that made it to the wikipedia page:

Datafolha - sample size: 12,800  - Lula +14
IPEC - sample size: 3,008        - Lula +16
Abrapel/Ipespe - sample size: 1,100 - Lula +13
Atlas - sample size: 4,500  - Lula +9.1
CNT/MDA - sample size: 2,002 - Lula +7.9

The economist polling aggregator had it at Lula +13 and CNN brasil had it at Lula +14 (picking out the two aggregators from sources you are likely to know).

https://en.wikipedia.org/wiki/Opinion_polling_for_the_2022_Brazilian_presidential_election

Lula won by 5.2%. Except for the one poll that had it at Lula +7.9%, that is probably going to be outside the margin of error (just to be sure about the Atlas poll, which was Lula +9.1% and could have theoretically published a large margin of error, their published margin of error was 1%).

That is a major polling miss. The top pollster listed surveyed over 12,000 people and was off by 9 points...I mean if you say "we think this is going to be result x but could be off by 9 points in either direction" you just gave an 18 point spread. What is the point of even conducting a poll with an 18 point spread? I bet I can get within 18 points of the republican / democrat spread in the 2024 presidential election and I don't even know who the candidates will be!

[Jacob]

Given how:

1) The populist reactionary right frequently combines unethical behaviour with accusations that their opponents engage in that behaviour.

2) The polling has been wrong recently, always on the side of underestimating the strength of the populist reactionary right.

3) The populist reactionary right frequently accuses their opponents of voter fraud.

4) The populist reactionary right has not show any evidence of being committed to democracy - the institutions, the process, nor the philosophy.

5) Where voter fraud has been in evidence it seems to be in favour of the populist reactionary right.

... I would not be entirely surprised if it turns out that it's not so much that pollsters are consistently wrong as it is that the populist reactionary right is engaged in electoral fraud that hasn't been detected.

[Barrister]

... I would not be entirely surprised if it turns out that it's not so much that pollsters are consistently wrong as it is that the populist reactionary right is engaged in electoral fraud that hasn't been detected.

So as mentioned, I am an interested observer from abroad when it comes to Brazil (and certainly much more so Brazil than any other country in Latin America), but not to the point where I can speak to the finer details of its voting system.

But you didn't word this in terms of Brazil.  You worded it in terms of "populist reactionary right".  Which I think means Trumpist/MAGA types, Orban, and a few other parties in Europe.

So of the two most obvious cases (Trump and Orban) - they certainly have undemocratic impulses!  Trump sent an armed mob to the capitol January 6, tried to get his vice-president to reject certain electoral votes.  Orban controls almost all media in Hungary, has control over the courts.

But what I have never seen is any credible evidence that either are falsifying vote totals.

So again - I can't speak at all to the quality of Brazilian political polling.  But it does seem like there's a persistent issue in western political polling.  I think this has been well-covered.  It seems to be a combination of people not having land-line phones, people not willing to answer calls from unknown callers, and a general mistrust of authority (in particular by right-wing voters).

[alfred russel]

Given how:

1) The populist reactionary right frequently combines unethical behaviour with accusations that their opponents engage in that behaviour.

2) The polling has been wrong recently, always on the side of underestimating the strength of the populist reactionary right.

3) The populist reactionary right frequently accuses their opponents of voter fraud.

4) The populist reactionary right has not show any evidence of being committed to democracy - the institutions, the process, nor the philosophy.

5) Where voter fraud has been in evidence it seems to be in favour of the populist reactionary right.

... I would not be entirely surprised if it turns out that it's not so much that pollsters are consistently wrong as it is that the populist reactionary right is engaged in electoral fraud that hasn't been detected.

This is just a recent analysis of the phenomena of the direction of polling misses, but I'd say a majority of thought seems to be that the direction of polling misses is not skewed toward the right (though it seems that way because of a few high profile cases).

https://fivethirtyeight.com/features/will-the-polls-overestimate-democrats-again/

This takes a long time to get to the punchline and is on a broader topic than this, but go to the link and start at point 1 of the article (there are some numbers to back this up and i'm selectively cutting and pasting below):

1. Polling bias hasn't been predictable historically
Our historical database of polls shows that there's not much in the way of consistent polling bias. Two cycles of a pro-Republican bias in 1998 and 2000 were followed by a Democratic bias in 2002. A fairly sharp Republican bias in 2012 reversed itself, and the polls were biased toward Democrats in both 2014 and 2016....

4. Polls haven't had a Democratic bias in elections without Trump on the ballot
As you can see in the table in the first point, polls did not have a systematic Democratic bias in 2018. That seems relevant, considering that was the most recent midterm.

Polls have also generally not had a Democratic bias in other elections in the Trump era when Trump himself was not on the ballot. They didn't have one in the Alabama Senate special election in 2017, for instance, or the Georgia Senate runoffs in January 2021, or in last year's Virginia gubernatorial race.

There have also been some races where Democrats have overperformed their polls, such as in last year's California gubernatorial recall election and in the 2017 governor's race in Virginia. But these errors don't tend to get as much attention from the media as those that underestimated Republicans....

5. Polls have been unbiased or underestimated Democrats in recent elections in 2022
Democrats have had a lot of success in elections since the Supreme Court's Dobbs decision — and importantly for our purposes, they've done as well or better than polls predicted in these races:

In the Kansas abortion referendum, the only poll showed the "yes" side — which would have enabled the Kansas Legislature to implement additional abortion restrictions — winning by 4 percentage points, but it lost by 18 points (!) instead.
Meanwhile, the only nonpartisan poll of the special election in Minnesota's 1st Congressional District showed Republican Brad Finstad winning by 8 points, but he actually won by 4, although an internal poll released by his opponent did underestimate Finstad's margin.
That said, polls of the special election in Alaska's at-large House seat were quite accurate in predicting both the first-round vote and that Democrat Mary Peltola would narrowly defeat former Alaska Gov. Sarah Palin if they were the final two candidates using Alaska's ranked-choice system.
Finally, of the five polls of the special election in New York's 19th Congressional District, which were from a mix of Democratic and Republican firms, all showed Republican Marc Molinaro winning by margins ranging from 3 points to 14 points. But Democrat Pat Ryan won by 2 points instead.

[Josquius]

There's pretty clear evidence trump tried this at least in Georgia.

Though I remember reading a piece the other week about the Brazilian voting system and the inevitable repeat of trump Post us election if bolsarnaro wins. There have been independent investigations there which have found no evidence of fraud.
But then the Lula trial was independent too so...

[alfred russel]

Probably my favorite example of totally shit polling...

2021 in Peru. In the first round, the current president Pedro Castillo (who is on the far left if it matters) never scored as high as 7% in polling. Polls within a month of the election had him under 5%. In the end, he was the highest first round vote getter with 18.9% of the vote.

https://en.wikipedia.org/wiki/Pedro_Castillo

[Admiral Yi]

It is an improvement and I seriously question how closely you actually follow Brazilian politics if you think Bolsonaro and Lula are remotely equal in how terrible they are.

Lula is corrupt to varying degrees.

Bolso wants to start a civil war and end democracy in Brazil.

+1

[DGuller]

It doesn't surprise me that the polling is hard, because the difficulty lies in your sample being selective rather than being small.  Sample error, or margin of error, is the trivially easy part.  The very hard part is identifying and properly correcting for bias in who responds to you, and even then your job is to aggregate what people are saying at the moment they're polled, not how they'll actually vote later.  If they lie in a systemic manner, or change their minds in a systemic manner, then that's not within your power to predict, and arguably not your job either.

[Jacob]

But you didn't word this in terms of Brazil.  You worded it in terms of "populist reactionary right".  Which I think means Trumpist/MAGA types, Orban, and a few other parties in Europe.

Yes, you took my meaning correctly.

But what I have never seen is any credible evidence that either are falsifying vote totals.

On Hungary, I'll defer to Tamas. My impression (not evidence) is that the Orban has enough fingers on enough levers that he could swing things if he needed to, but that he hasn't had to (though there's no way to actually ascertain that either way I don't think).

On Trump, all I've seen is a few individual MAGA types casting fraudulent votes, and I think at least on case of a lesser GOP official attempting to improperly access voting machines somewhere. Which, I agree, is not credible evidence that they are falsifying voting totals. I mean, there is the attempt to overthrow the 2020 presidential election, and Trump's call to find "the necessary votes" which I think is now on the record. But yeah, it's a suspicion, not actionable evidence.

[Barrister]

An Amicus brief filed by The Onion in a case touching on the protections and limits of parody.

I. Parody Functions By Tricking People Into
Thinking That It Is Real.

Tu stultus es. You are dumb. These three Latin
words have been The Onion's motto and guiding light
since it was founded in 1988 as America's Finest News
Source, leading its writers toward the paper's singular
purpose of pointing out that its readers are deeply gullible people.
The Onion's motto is central to this brief for two
important reasons. First, it's Latin. And The Onion
knows that the federal judiciary is staffed entirely by
total Latin dorks: They quote Catullus in the original
Latin in chambers. They sweetly whisper "stare decisis" into their spouses' ears. They mutter "cui bono" under their breath while picking up after their neighbors' dogs. So The Onion knew that, unless it pointed to a
suitably Latin rallying cry, its brief would be operating
far outside the Court's vernacular.

or

Here's another example: Assume that you are
reading what appears to be a boring economics paper
about the Irish overpopulation crisis of the eighteenth
century, and yet, strangely enough, it seems to advocate for solving the dilemma by cooking and eating
babies. That seems a bit cruel—until you realize that
you in fact are reading A Modest Proposal. To be clear,
The Onion is not trying to compare itself to Jonathan
Swift; its writers are far more talented, and their output will be read long after that hack Swift's has been
lost to the sands of time. Still, The Onion and its writers share with Swift the common goal of replicating a
form precisely in order to critique it from within.

This is the fifteenth page of a convoluted legal filing intended to deconstruct the societal implications of
parody, so the reader's attention is almost certainly
wandering. That's understandable. So here is a paragraph of gripping legal analysis to ensure that every
jurist who reads this brief is appropriately impressed
by the logic of its argument and the lucidity of its prose:
Bona vacantia. De bonis asportatis. Writ of certiorari.
De minimis. Jus accrescendi. Forum non conveniens.
Corpus juris. Ad hominem tu quoque. Post hoc ergo
propter hoc. Quod est demonstrandum. Actus reus.
Scandalum magnatum. Pactum reservati dominii.
 See what happened? This brief itself went from a
discussion of parody's function—and the quite serious
historical and legal arguments in favor of strong protections for parodic speech—to a curveball mocking the
way legalese can be both impenetrably boring and belie the hollowness of a legal position. That's the setup
and punchline idea again. It would not have worked
quite as well if this brief had said the following: "Hello
there, reader, we are about to write an amicus brief
about the value of parody. Buckle up, because we're
going to be doing some fairly outré things, including
commenting on this text's form itself!"

Fantastic.

https://www.supremecourt.gov/DocketPDF/22/22-293/242292/20221003125252896_35295545_1-22.10.03%20-%20Novak-Parma%20-%20Onion%20Amicus%20Brief.pdf

[Valmy]

Holy shit valmy, they were wrong and I can't even fathom why you are upset about this....

Sure they were wrong as to the exact percentage but was it the expectation prior to 2016 that polls nail it so exactly? I can certainly recall many elections prior to 2016 when the polls were wrong. I am just curious why this is considered some kind of important new trend.

[alfred russel]

I am just curious why this is considered some kind of important new trend.

Seriously? You jumped into defending the polls which is a strange way to express that curiousity...I pointed out that the largest sample size poll had Lula winning by 14% and if you have a 9% margin of error that makes the poll completely meaningless. Anyone knowledgeable about Brazil should have been able to guess the results to within 18% without the effort of getting responses from over 12,000 people (and considering response rates they probably asked a multiple of 12,000 people).

There has been a lot of literature on recent polling challenges, but if you want something right at hand...that link I gave above from 538 has the polling errors for 4 types of elections (president, governor, senate, and house) in the most 12 recent cycles - that is 48 total data points. Since 2010, in 15 of the 18 data points the results of the election have exceeded the average for the type of election. 2020 the biggest miss since 2000 in every category. In the US context, it does look like polling quality is deteriorating.