Maybe in some form but 538 was founded in 2008. It’s also changed owners since then. It started independent, was bought by the NY Times and is now owned by ESPN.
So, is the claim that he hasn’t introduced bias in his predictions to show a closer race before, but now he is?
That can also be tested (although obviously not until we have new election results). Yes, it’s possible, at any given point, that Nate Silver has decided to throw his principles overboard, and we won’t know it for a while. But if that’s the claim, I hope someone makes it clear and then shows us how right they were after the election.
OK, I’ve not seen this sufficiently explained but I am sure there is an article out there if someone would kindly link: If Silver went 50/50 in predicting State wins doesn’t that by itself show he’s lowballing the odds? By way of illustration: if he put the odds at 90-10 in each state, doesn’t that mean he should have been “wrong” in around 5 states?
One has to remember that 538 did predict how dismal the Democrats would perform in the last mid-term election. IMHO they are still very fair with the data and there were no favors made to the Democrats in 2014.
I don’t understand this response. I was merely pointing out that he hasn’t been predicting elections for over a decade and he hasn’t been under the same employer for the whole 8 years that he has. I never mentioned anything about a party bias.
Nonsense. Even if he had not started the campaign with high negative ratings and then continued to make them worse with his various offensive comments, and even if he put in those great debate performances, he would still face an uphill climb.
The problem with any discussion of how close or not-close the race is today is that there is no proof of anything until November.
There is a pollster, I think it’s Rasmussen, that was an outlier in 2012, but the nearer it got to November, the closer the numbers came to the pack, so that the “final” poll was not an outlier at all. And then the election happened, and that pollster was right.
Silver could basically be doing exactly that, but with the predictions that are trying to account for uncertainty. He could be building in a bias toward an electorate that is completely up in the air. Then, as we get closer to November, he can edit out that swing and end up with a prediction that’s right.
Texas paper that has endorsed the Republican candidate since 1968, won’t endorse Trump:
Some quotes:
Now I know newpaper endorsements aren’t the thing they used to be, but when a paper with nearly a 50-year history of endorsing the Republican nominee takes a look at this one and says, “nope, I’m gonna pass”, I think it’s worth mentioning.
Note, they did not endorse Hillary per se, but they sure didn’t give the Trumpsters any good news…
He has done those kinds of post-analyses in the past, and he does in fact miss calls with fairly appropriate frequency. Like, if you look at all the races that he calls 60/40, the candidate with the 60 wins about 60 percent of the time.
It’s possible, but this would be hard to hide. If he is biasing earlier models toward closeness, then we should see a trend near the end towards a more unbalanced race. And, in particular, it should be unsupported by any of the input data he claims to use. While his exact model is a secret, he regularly discusses the inputs (polls, economic data, etc.) he uses, and the philosophy he uses to rate them.
We can continue to speculate about ways he might cook the books, but there’s simply no evidence that he’s doing so so far. And most of the ways he might do so would damn him in retrospect. So far he’s built his increasingly-popular media empire on painstaking accuracy. What makes people think he’ll throw away his good name for, what, an increase in page view for a couple months in one election?
They announce the endorsement tomorrow. Wanna bet it will be Trump anyway? Maybe they have endorsements written up for each of the four depending on what the reaction is to the savaging of Trump.
After all, Ryan and McConnell are thinking the same thing and still support him. McCain, after being insulted(and the military), supports him.
No. I’m not talking about changing the poll numbers, only the assumptions made about how variable they can be.
One person looks at a race where the D is ahead by 5 and says “This is a huge lead.” Another person looks at the same race and says “This is an okay lead, but the other guy can come back.” And then build a model that ensures that “the other guy can come back” happens often enough to justify that punditry. As the race goes on, the model changes to account for less time to make up the difference.
There is no way at all to “out” this type of modeling, since the assumptions being made cannot be falsified.
For starters, it was my claim, not CarnalK’s. He can’t defend my thoughts on this. Just FYI.
Secondly, I have no doubt that Silver’s model passes mathematical purity tests of powers beyond my poor comprehension. However, the narrative on their podcasts and articles is still akin to the “horserace” narrative that drives clicks and stories, making it appear that Trump is a mere half a body behind Clinton rather than being 12 furlongs back.
“Oh, no! The horse broke his leg/has no field staff!”
“There’s still plenty of time to catch up and win, though!”
If they start calling this one done, clicks and ratings will start to drop, only bouncing back to today’s level on November 9th and only in the case that Trump wins. It’s just a basic economic fact which impacts the families of Nate Silver and his staff… and therefore they can’t make that call.