Categories
The Battle of Trafalgar – politicalbetting.com
The Battle of Trafalgar – politicalbetting.com
So… I spent 30 minutes on the phone this morning with Robert Cahaly, the Head of The Trafalgar Group which PBers will recall as the pollster who called Michigan, Pennyslvania and Wisconsin correct back in 2016.
0
This discussion has been closed.
Comments
Sounds like if they are right it could be pure luck. Which in fairness pollsters of more traditional (or at least clear) methodology sometime seem to end up like.
All headers should end with 'I could, of course, be wrong'. Not only will it cover all authors completely, it will identify, if it is anonymous, that it definitely wasn't HYUFD.
I could, of course, be wrong.
(edit) But Robert’s efforts go above and beyond, and are greatly appreciated.
Someone was asking about the possible legal consequences for Trump should he lose.
Here’s a long (but almost certainly not exhaustive) list:
https://edition.cnn.com/2020/10/17/politics/trump-election-legal-reckoning/index.html
Certainly don't seem plausible. Besides potentially getting lucky with 2 states in 2016 which seems to get focused on by their fans, most of their predictions were very, very wrong but those get ignored.
An old story: make a dozen outrageous predictions, when one comes true claim to be a sage because you predicted that one, ignore everything else you predicted that was wrong.
But I wonder if those who do get it wrong, will have it brought up 10 times a day or will they be ignored? Will they pretend they never made those predictions? Who knows, only time will tell.
Trafalgar, or Mr Cahaly to be precise sounds completely bogus as a pollster. He applies about as much scientific analysis as is associated with completing the football pools. So in reality it is little more than a finger in the air test. Like the football pools every few weeks one might get two or three draws right and on a very lucky week, four, and if one is especially fortunate, once in a blue moon five.
Go bold, or go home
2020 will be like 2019 barring a shock. Even with a late shock it may still be like 2019 given the volume of voters already going to the polls.
- Think they are outsiders but are actually just a bit bonkers
- Have completely taken over a political party which was hostile to them
- Repeated dodgy views and behaviours on racial issues
- Attracted a cult of fanatical loyalists but increasingly disgust moderate swing voters
- Old white men who were born into wealth but like to pretend they are "of the people"
- Lost the popular vote (can shortly add "twice" to this)
- Both supported by Piers Corbyn and (intermittently) David Duke
- Interfere in the selection/primary process in attempt to remake party in their image
- Incapable of admitting when they are wrong, or have failed
- Claim evidence of their wrongdoing is "fake news" or "smears" even when it has been photographed or taped
I'd say there are a few similarities.
What would the cost of such a large up to date/contactable data set even be? Cos you're gonna need a big block to over come the response rate And why is such a data set from a specific company not skewed in the first place it must have come from some where?
And
Why do the polled have to be any more truthful than to traditional polling companies?
What is the methodology difference from the perspective of the chronically untruthful? Isn't their reaction to one polling agency on the phone much the same to any other? Why does Trafalger put such cynical voters at ease does Robert tell each pollee that their phone call is a safe space?
As a recovering former Corbynite (if I may call you that) can you recognise now the similarities between Corbyn and Trump that have been pointed out? And the similarities between his hardcore followers which you now distance yourself from and Trump's hardcore followers?
Would be better if everyone focussed on my tip of Labour majority in NZ...
https://twitter.com/atrupar/status/1317979259764461568?s=21
Think "it was all about how to make it as anonymous as possible" is excellent way to describe way Mr Cahaly conducts his business. Sub rosa save for the PR. Right out of the Lee Atwater playbook.
If you put on a blindfold and turn around, every once in a while you will still pin a tail on a donkey in the correct place. If you take 100 pollsters and try the same thing, one of them will pin it on correctly and you forget the 99 who didn't. It's like today's Daily Mail front cover reporting on the astrology who 'predicted' the Covid pandemic but neatly forgetting the 99 who didn't.
It's a popular misconception that the 2016 polls got it wrong. In fact the final 2016 result with Clinton winning the popular share by c. 2% was pretty much within the margin of error of final polls: https://en.wikipedia.org/wiki/Nationwide_opinion_polling_for_the_2016_United_States_presidential_election#Polls_conducted_in_2016
A few overstated her lead but by and large they weren't far out.
But, besides, the 2016 meme was totally different. Trump was galvanising disaffected rust-belt voters. Much like the disaffected white working class in this country on the Brexit vote: it was about outsiders taking back control.
This time the reverse is happening. Trump is the swamp. There's no miracle this time around for him.
Furthermore, as Mike has noted, there is an unprecedented early voter turnout going on.
I know a lot of people over here don't go deeper than the Presidential race but we should do: it's really interesting as well as kind-of vital to the next four years.
The world is changing. The flirtation with right-wing bombast has come to an end.
Not least because this one guy is the one making the phone calls, and he clearly can’t stick to a topic without demonstrating his bias.
https://science.sciencemag.org/content/368/6498/1422.full
Aerosol transmission of viruses must be acknowledged as a key factor leading to the spread of infectious respiratory diseases. Evidence suggests that SARS-CoV-2 is silently spreading in aerosols exhaled by highly contagious infected individuals with no symptoms. Owing to their smaller size, aerosols may lead to higher severity of COVID-19 because virus-containing aerosols penetrate more deeply into the lungs (10). It is essential that control measures be introduced to reduce aerosol transmission. A multidisciplinary approach is needed to address a wide range of factors that lead to the production and airborne transmission of respiratory viruses, including the minimum virus titer required to cause COVID-19; viral load emitted as a function of droplet size before, during, and after infection; viability of the virus indoors and outdoors; mechanisms of transmission; airborne concentrations; and spatial patterns. More studies of the filtering efficiency of different types of masks are also needed. COVID-19 has inspired research that is already leading to a better understanding of the importance of airborne transmission of respiratory disease.
https://edition.cnn.com/2020/10/18/politics/what-matters-october-18/index.html
Lara Trump defends president's rhetoric after 'lock her up' chants about Whitmer at rally: 'He was having fun'
https://thehill.com/homenews/administration/521585-lara-trump-defends-presidents-rhetoric-at-rally-after-lock-her-up
More likely, his approach generates shy democrats, particularly those living in republican areas who are continually bombarded with the sort of rubbish I was listening to on the Mike Gallagher show when I was in the US last year.
538 did very well in 2016 and were within margin of error for the result. In multiple states Trafalgar were simply not.
A point add to the many other comments sceptical of Trafalgar's methods. Lets suppose that Trafalgar has hit on a great new polling method, a method which turned out to accurately predict results in the swing states in 2016. If this new polling method is based on a sound new methodology, then how come none of the other polling companies have switched to using it? These are private companies most of whom want to be judged on their presidential election polling accuracy. That they stick with other methods suggests that the Trafalgar methodology is very shaky.
Ralston is the man who knows Nevada and early voting. When Ralston calls the state his word is law.
The polling looks really good for Biden, not least in its consistency, but I think reading through the entrails of early voting returns tells you very little.
But I'm afraid early voting returns need to be taken with a pinch of salt. It reminds me of the morning of 12th December when we were told, and shown pictures of, a massive youthquake.
I expect Joe to win Nevada handsomely but the conservative Trump Rump will still show up on the day.
But much like Old North State politics blog for North Carolina I thought people would appreciate a local expert who does election analysis.
Use at your own risk.
People grasp for it as the only information we have during the increasingly febrile runup to a critical election.
We see the same at UK election time, with endless posts about snippets of information supposedly derived from postal vote verifications, which are always rubbish.
There is no post collection weighting of results, which seems odd, because given admittedly low response rates, it would seem necessary. There’s also no “spiral of silence” type adjustment.
But the end of their 2016 polling they said
Survey results undergo a weighting process to ensure the sample is comparable to the demographic breakdown of 2012 Colorado general election participants. Final results are based on these weights in order to address non-response bias. In addition, the final published ballot test is a combination of survey respondents to both a standard ballot test and a ballot test gaging where respondent's neighbors stand. This addresses the underlying bias of traditional polling, wherein respondents are not wholly truthful about their position regarding highly controversial candidate
They both cannot be true.
Which if you think about it is the same fiasco as we saw in March. Don't go to the pub spake the ministers. "what about our livelihoods" say the hospitality trade. "what about it" says the government.
Fuck business vote Tory. Again. How is it they haven't learned a single thing from the first phase of this pandemic?
The GOP ratio keeps to a narrower band than the Dems, Clinton had a particularly poor ratio in 2016 and still won even with Trump hitting the top end of the GOP ratio.
Nevada doesn't have a "Southern Democrat" problem when it comes to voter reg analysis
In Texas nearly half of 2016's turnout has already voted. That's not insignificant. Sometimes people can read to much into it, or see what they want to see, but it's not necessarily useless.
I think as a rule electoral returns reports mean more when people start getting downbeat about their own prospects from them rather than seeing or saying what they want to see.
https://www.leicestermercury.co.uk/news/leicester-news/businesses-threatened-by-council-breaking-4617328
Everyone defers to him on his knowledge of races. He called Harry Reid winning when everyone was sure he was a goner. Reid did win against the tide in 2010, the only bright spot for the Democrats that night.
If there are any betting markets on Nevada EC votes, you'd be mad to go against Ralston.
But that is based on September registration figures. I need October's before I make by (Ralston willing) call.
The unelected Lords should have their say but after ping pong respect the elected Commons decision.
For example, Pennsylvania. Latest Trafalgar is Biden +2, but there's also an Insider Advantage poll with a sample size of 400(!) which is only Biden +3.
The polls and now quite a bit of anedotal evidence are suggesting a shellacking for the GOP on a scale rarely- if ever- seen before. Even allowing for a hard core of Republicans voting solidly in the red states, we can`t be looking at more than a third of the EC going Trump, and there could be some big name casualties in the Senate.
It´ll be interesting how Trafalgar explain that ex post facto. Probably have to rename the dog to Villeneuve...
Just a question. I'm sure online panels work as a polling process, but I am puzzled as to how. Is it not inevitable that online panels are biased towards the sorts of people who like being on online panels, biased towards those who have some sort of interest in politics and also skewed by the fact that just being on a panel may affect how much attention you give to the issues as compared with a constantly changing random sample.
This is not a proper polling operation. If it gets anywhere near the right answer it will be by a fluke. It wouldn't change the nature of polling because his results could not be copied, not least because nobody can really know exactly what he is doing. For betting purposes I would do pretty much as 538 does - chuck the figures in the mix but give them less weight than the other proper firms.
As for the betting, once more we have an overnight shortening of Trump's odds, which makes me think much of the support is from abroad. They tend to ease back during the day. It may be someone priming the odds; that's happened before. I suspect however it is simply over-exuberant Trump supporters unacquainted with or dismissive of the facts.
There were relatively few polls over the weekend and the most recent, IBD and CBS (who are decent pollsters, no doubt), did suggest a slight weakening of Biden's support. We will know soon the extent to which other pollsters confirm this. If they do, I would attribute it to the kind of tightening of the polls near election day which is common in elections around the world. I would not attribute it to Trafalgar knowing something others do not.
Finally, early voting returns do not tell you everything but they do say something. They're a bit like 'runs on the board' in cricket. You don't know how good they are until both sides have batted, but on balance you'd sooner have them than not. I think they are encouraging for Biden, but no more than that.
Right, gotta walk the dog. Catch you all later.
I have my finger on the pulse.
Only you with your fantastic politeness and patience could have pulled this off.
--- "I'm sorry, you've lost me there..."
Perhaps a combination of that and glorified focus group to test Republican lines or target electoral groups ?
Like Robert, I'll believe they are some kind of pollster should they end up right again. You can get lucky once; twice in the same manner is unlikely.
In 2016 there were 577,679 Registered Dems and Clinton got 539,260 votes
488,861 Registered GOP and Trump got 512,058 votes Gore was not popular in Nevada!
And the figures for this will be very interesting:
States brace for surge of voter registrations as deadlines near
https://www.politico.com/news/2020/10/18/voter-registration-deadlines-online-glitches-429632
Anyone betting on turnout ?
In a lightly traded market, Biden is favorite to win Nevada at odds of 2/5.
Just to reinforce the point, in terms or raw outcomes (deaths & economic hit), we are near the worst in the world.
https://twitter.com/ianbremmer/status/1317924980659785728
Regarding the first point, I expect that weighting is done by comparing the results of those recruited to online panels compared to those polled in other ways, so any systematic differences in those who would like to be in online panels can be accounted for.
The third point is interesting, would be fascinating to see whether people in online panels become more aware of the things they are asked about over time. Again I expect you can compare new and old recruits/panels to tease out these differences, although that may be too subtle to have been done so far. Intellectually, that third question is intriguing.
https://twitter.com/FinancialTimes/status/1318084481241223170?s=20
Though in 2016 while they were the only pollster to correctly have Trump ahead in Michigan and Pennsylvania not a single pollster had Trump ahead in Wisconsin but he won all 3 ( Trafalgar did not poll Wisconsin then and is this time as even Trafalgar did not expect Trump to have a chance in Wisconsin in 2016, yet Trump won it by the largest margin of the 3).