Chrkeller said:
the-pi-guy said:
Thanks for bringing in actual data. Chrkeller was arguing that companies don't like to lose money. Mistakes can still be made, even on the industrial scale. Bad data can make misleading conclusions. Even good data though, can be interpreted completely incorrectly. Especially when you don't have all the facts. |
Bad data like Trump being under polled in 2016 and 2020? Just because you don't like that data doesn't mean it is bad data. The data has it being tight, with an edge to Trump. You shouldn't backhand jab the messenger. And yes, professional betting firms aren't in the business of losing money and I'm certain they understand data better than the random folks in this thread. |
No, people are not dismissing the polls because they don't like them, they are dismissing them because they are bad.
Atlas Intel is included in the fivethirtyeight average. They say that Trump will win 45.7% of the urban vote and about 30% of the black vote. That is absurd. According to their sample, 88% of their respondents voted for Biden in 2020, and 96.9% voted Trump. Of course, Biden won the popular vote in 2020, so their sample clearly not representative and skewed towards Trump.
https://www.atlasintel.org/poll/usa-national-2024-1710
Activote is a poll that I believe is not really sampled. Whoever clicks the link to do the poll does it. This leads to wonky results. For example, they have Trump winning by 1. They also have a sample that when unweighted was 64% male. They did weigh it to 54% female, which is about what the voting population was last year, but the fact that they had such a skewed sample in the first place makes it questionable. More problematic is that even after weighting, their sample was 32% urban and 32% rural and 37% suburban. In reality, rural voters at best make up 20% of the population. As with Atlas, they have Trump winning about 30% of the black vote, which would be a pretty crazy result. Even crazier, they have Trump winning 19% of registered Democrats which is pretty wild. This was 5% in 2020.
https://www.activote.net/trump-extends-lead/
Beacon, from Fox News, again has Trump winning 29% of black voters. Is this impossible? No, but Trump almost tripling his support among black voters from 2020 is a strange result. This poll also says Trump will be winning 46% of the 18-34 year old vote, up from 30%. Meanwhile, Trump is winning among college educated whites with 51% while he lost this group by 13% in 2020. What is stranger is that this number has just about flipped since September.
https://www.foxnews.com/official-polls/fox-news-poll-trump-ahead-harris-2-points-nationally
This is what I can see from fly by night pollsters who make their crosstabs available. Either there has been a massive surge in Trump support among the youth, black people, and even among Democrats, or the polls are wrong. Personally, I think the latter is far more likely. Trump is still Trump except older, more sluggish, and more orange. It's hard to see why there would be such a sudden surge.
You might say that immigration and inflation are so bad, or perceived as so bad, that they would cause lots of people to suddenly turn to Trump. Yet, the problem with that is that just about a month ago, these same polls were showing much better numbers for Harris. And I don't think these issues are getting any worse. Plus, there are countervailing factors that would likely boost Harris compared to Biden in 2020, such as Dobbs, January 6th, Trump's continued deterioration, and the fact that she is not 80 years old.
We shouldn't dismiss polls that give us news we don't like, but there is very good reason to be skeptical of many polls.
As for the betting markets, you are correct that they are not out to lose money. However, you are wrong that could somehow lose money. Polymarket at least, and I imagine the others, make money by charging a transaction fee when bets are placed. If I place a 5 dollar bet and you place one, they take 25 cents from each of us, and the 10 dollars goes in the pot. Whether I win or you win, they get 50 cents. It doesn't matter how well they can analyze data because they are not analyzing data. If they were wanting to make as much money as possible, I believe it would probably behoove them to keep the results right around 50-50.