onejayhawk
Afflicted with reason
Remember that control includes chances of 50/50 plus the White House. Outright control is looking good but not as good.
Don't get so excited over small fluctuations. Random chance plays a factor as well.
J
[RD] US 2016 election: Poll watching thread
- Thread starter dutchfire
- Start date
Don't get so excited over small fluctuations. Random chance plays a factor as well.
J
Gori the Grey
The Poster
- Joined
- Jan 5, 2009
- Messages
- 10,998
Unless they go against Hillary, in which case they are an indisputable sign of an unstoppable trend that will lead inevitably to her utter demise.
But (not "outright") control factors in the relative chances of winning the presidency. I don't understand the insistence on making the distinction of outright control, although of course Democrats would prefer a faction whose control isn't endangered by a single defector or otherwise lost seat.
Joe Lieberman (may he burn in she'ol) proved that one defector can be catastrophic... For those that don't remember, at the behest of his insurance-company masters, he singlehandedly torpedoed the public option from the final version of the ACA. He is literally the reason that Obamacare has no public option included, and in that sense, probably partly responsible for the recent rate hikes, which would have been less likely in an environment where people could simply dump private insurance in favor of Medicare.
Trump's surrogates are doing a huge disservice in confusing people about the nature of polling and statistics. I'm starting to post one-sentence answers in the youtube comments. You can only change people's minds with mindworms. But it needs to be done. We're seeing a crop of people who now think they've figured out a 'gimmick' in an entire branch of mathematics.
metalhead
Angry Bartender
- Joined
- Apr 15, 2002
- Messages
- 8,031
Every time a poll comes out that is lower than the RCP average you come on here talking about how Trump has regained momentum. Perhaps you should heed this advice yourself, rather than hypocritically dispense it to others.
Minor fluctuations only help the underdog this early in the campaign. It will be devastating to Clinton when people start paying attention to the election and we move out of preseason.
onejayhawk
Afflicted with reason
I missed an interesting detail the first time through the RCP averages the other day. The Gravis poll is of registered voters. At this point, almost all polls are likely-voter polls because they are thought to be more predictive. So, it is interesting because it's unusual, though not unique (Economist/YouGov).
More generally, this race is clearly not like 2012. That year the last two weeks were all about President Obama. This year, neither side has any clear momentum for the last dozen days. Nate Silver did an article on the general question of whether the race was tightening. His answer, "probably."
http://fivethirtyeight.com/features/election-update-is-the-presidential-race-tightening/
J
More generally, this race is clearly not like 2012. That year the last two weeks were all about President Obama. This year, neither side has any clear momentum for the last dozen days. Nate Silver did an article on the general question of whether the race was tightening. His answer, "probably."
http://fivethirtyeight.com/features/election-update-is-the-presidential-race-tightening/
J
I agree that at this point, LV polls are probably more appropriate than RV polls. I said as much several weeks back, when RV polls were still the best measure.
metalhead
Angry Bartender
- Joined
- Apr 15, 2002
- Messages
- 8,031
Gravis is totally useless. Not only are they doing RV polls, their horse race forces people to choose one or the other. Anyone who knows anything about polling will tell you that results gathered in that manner are totally unreliable.
That "probably" includes the caveat that Clinton's projected vote share is still increasing, and that Trump's larger increase in projected voter share hasn't moved his odds of winning very much (13% -> 15%). Trump's improved polling is likely due to Republican's coming back to him now that his media coverage isn't quite as disastrous as it was for a few weeks there.
stfoskey12
Emperor of Foskania
Trump now has a >20% chance of winning according to the 538 polls plus forecast. That's quite concerning. It's a 3.4% drop just today. At this rate it will be a 60-40% Trump odds of victory by election day. While the decline may not be that fast, I like a far greater chance of certainty than only 60 or 70%. I can't tell you how many times I've seen a 60-70% chance of snow, or even 80%, and then I see it not happen.
onejayhawk
Afflicted with reason
Not everyone. In fact, not most.
That was not a caveat. That was a central theme--Trump may be improving faster but both are improving. Regardless, whether the race is narrowing is unclear.
J
It's so funny how all Trump has to do is not be in the news to improve his polls.
a) chance of a polling error in Trump's favor
b) chance of Trump doing something to persuade more voters
Both of these things need actual reasons to happen, I just don't see it.
It's important to understand the difference between those probabilities. For weather, it mostly reflects the uncertainty inherent to a chaotic and unpredictable system. Here, there are mostly two factors:
a) chance of a polling error in Trump's favor
b) chance of Trump doing something to persuade more voters
Both of these things need actual reasons to happen, I just don't see it.
Last edited:
Trump is now consistently slightly ahead in Ohio again over the last six polls that have been released there, but 538 has its odds of being the tipping-point state at only 5.9% or sixth place. All three of Ohio, Iowa, and Arizona are roughly 50-50 shots now; these are the most Trump-leaning of the states that can reasonably be called swing states, whereas Georgia and states redder than it would only flip in situations where Clinton wins by a large margin. The next tier consists of NC, FL, and NV, where she's at about 2-1 odds to win.
At this point it does actually look like Trump's post-debate collapse is over and he's slowly regaining again, in yet another Trump Cycle. It would be unlikely but not impossible for the polling to revert to the 272-266 case from just before the first debate, where Clinton goes back to winning with no margin of error by carrying PA, NH, and the Upper Midwest, where she has been leading consistently for the whole election. I'm sure that they have more damaging things that are scheduled to be released next week so as to stop this cycle short and bring Trump back down again, but ATM we're back to being a little too close for comfort.
At this point it does actually look like Trump's post-debate collapse is over and he's slowly regaining again, in yet another Trump Cycle. It would be unlikely but not impossible for the polling to revert to the 272-266 case from just before the first debate, where Clinton goes back to winning with no margin of error by carrying PA, NH, and the Upper Midwest, where she has been leading consistently for the whole election. I'm sure that they have more damaging things that are scheduled to be released next week so as to stop this cycle short and bring Trump back down again, but ATM we're back to being a little too close for comfort.
Right, but they all want to believe that. Statistical products can be fiddled with unintentionally by any variety of methods, and I strongly believe that in particular academic and liberal media (e.g. NYT) sources are going to demonstrate a pro-Clinton bias, particularly if those products are new. I trust 538 largely because I trust the people there are mostly interested in producing an accurate prediction and have a decently long record of doing this well.
That said, if the current polling numbers hold completely steady, the 538 odds will shoot up past 90% in the last few days of the election because their algorithm is giving a prediction based in part on the odds of unexpected events in the future. In the last few days of the campaign, those odds fall rapidly to 0.
I really don't believe 99%, in particular. The history of polling contains enough misses in the polling average by wide enough margins that there's no way that 99% Clinton is plausible, no matter how much data is aggregated. The 99% number is probably attained by some sort of error akin to the one by which the housing bubble collapsed, or Long-Term Capital Management before it: they likely assume too little correlation between state polls, and/or assign too small a value to "black swan" events on the tails of their probability distribution, and/or don't take into account herding behavior by pollsters, which was a major reason that polls in 2014 missed.
That said, if the current polling numbers hold completely steady, the 538 odds will shoot up past 90% in the last few days of the election because their algorithm is giving a prediction based in part on the odds of unexpected events in the future. In the last few days of the campaign, those odds fall rapidly to 0.
I really don't believe 99%, in particular. The history of polling contains enough misses in the polling average by wide enough margins that there's no way that 99% Clinton is plausible, no matter how much data is aggregated. The 99% number is probably attained by some sort of error akin to the one by which the housing bubble collapsed, or Long-Term Capital Management before it: they likely assume too little correlation between state polls, and/or assign too small a value to "black swan" events on the tails of their probability distribution, and/or don't take into account herding behavior by pollsters, which was a major reason that polls in 2014 missed.
The momentum isn't good. Trump might not have much dirt left, and people seem to have short memories. And wikileaks gives headlines that will tend to be negative due to the nature of sensationalism. Melania might turn on Donald or there might be another gaff. But his surrogates aren't having to spin his words into either false beliefs or something more reasonable
The polls tend to lag behind the news by a couple days. We had a couple Wikileaks (on Clinton only, of course). Plus the Obamacare hikes, which I can't believe went out before the election. Arizona was one of the hardest states hit by the Obamacare.
However, we now have Trump accusing Texas and saying we should just cancel the election and declare himself the winner. That'll p*** off Texas election officials (especially the pro-Trump ones--bad idea) and many undecided voters not interested in turning the country into Banana Republic. That should take a couple days to digest.
Barbara Corcoran also came forward as a Trump accuser (#13, I think?), and it's not like the millionaire entrepreneur is just interested in the notoriety. I doubt that would sway the polls that much, because that requires someone to examine the accusations closely. People voting Trump now aren't doing that--they either just don't care or think they're all lying. Barbara Corcoran has absolutely no reason to lie (let alone in syncopation with 12 others).
However, we now have Trump accusing Texas and saying we should just cancel the election and declare himself the winner. That'll p*** off Texas election officials (especially the pro-Trump ones--bad idea) and many undecided voters not interested in turning the country into Banana Republic. That should take a couple days to digest.
Barbara Corcoran also came forward as a Trump accuser (#13, I think?), and it's not like the millionaire entrepreneur is just interested in the notoriety. I doubt that would sway the polls that much, because that requires someone to examine the accusations closely. People voting Trump now aren't doing that--they either just don't care or think they're all lying. Barbara Corcoran has absolutely no reason to lie (let alone in syncopation with 12 others).
