You cited Nate Silver so I thought his views on the matter would be of interest. Incidentally, I do not pretend to understand the calculations converting margin of error into level of confidence, which Nate Silver considers as well in this article. I am however struck by his conclusions:
it didn’t take any sort of extraordinary, unprecedented polling error for Trump to defeat Clinton. An ordinary, average polling error would do — one where Trump beat his polls by just a few points in just a couple of states — and that’s the polling error we got.
You read that right. Polls of the November 2016 presidential election were about as accurate as polls of presidential elections have been on average since 1972.
I am confident that you are fully familiar with Nate Silver's thinking but I believe others on this thread who reflexively reject polls when they are unfavorable would find his article instructive.
Finally, here is Silver's bottom line The media narrative that polling accuracy has taken a nosedive is mostly bullshit
I am confident that you are fully familiar with Nate Silver Nate Silver wants us to believe that Trump beat his average poll amounts not just once or twice, but hundreds of times. Citing the average is misleading when you consider the fact that there were many, many polls all wildly incorrect.
Furthermore, it is a matter of absolute fact that these polls oversampled women and minorities, sometimes by 10 or even 15%. All of this was discussed at length at the time on this forum, with the numbers showing Democrats expecting turnout equal to or even better than the Obama years.
Your cheerleading of these polls is silly in light of that.
I am familiar with Nate Siver to this extent. He had Clinton with a insurmountable lead as far as the numbers went oif anywhere from 6-12 points and on these polls he based his probability factor of Clinton winning in the 90+% range right up until election day when he suddenly dropped that 90+ probability of Clinton winning down to a tad over 70..still a virtual lock on winning. For this he was attacked by crazed liberals for even DARING to insinuate her chances may have decreased.
The thing is if something within the MOE could cause a swing from 90+ Clinton will win all the way to Trump actually winning and by some 80EC votes why wasn’t this even mentioned before and why was Clinton always shown as the leader even if within the MOE?
Side note-The LAT/Dorsife daily poll of that race showed Trump much, much closer to Clinton and on several occassions actually leading her and the LAT caught holy hell for printing that poll every day.
It is easy to formulate excuses as to why they were so wrong rather than simply say we blew it and did not/could not account for the “silent” Trump supporter?
My guess is today there are more silent TYrump supporters than last time and the pollsters will have no more success identifying them this time than they did last time.
Yes, I did mention Nate Silver, but I only did so because he and I came up with the conversion formula at the same time (I can tell by examining his numbers and the timing of his appearance on the national stage). I had posted that fact many times in prior elections, but not in my response to you. He has the resources of the NYT, and I do not. It was his job to examine polls, I was working in corporate America in my "day job."
That said, others were right to call out the polling lookback as flawed because they only focus on the final polls and conveniently ignore the breathless hyping of Democrat landslide victories that suddenly evaporate after the Last Monday in October.
In Football, the cliché is that "the game was much closer than the score shows."
I think what we saw in 2012 and 2016 was the same thing: the race was much closer than the polling showed. I don't believe the race tightened only in the last week. The rally turnouts on both sides belie this belief. Trump was strong all throughout the general election, while the polls all went the other way. Trump was filling up arenas and Clinton was absent, yet she had the +10 polls in September and October.
To show how a close election can go either way with smart targeting, let me repost my 2014 post mortem of the 2012 election between Obama and Romney, where Obama won 51.1% to 47.2%, or 332-206 Electoral College votes:
The election came down to swing states, and to specific precincts in those swing states. So what happened in 2012?
Colorado went 51%-47% for Obama. Denver alone voted 209,759 to 69,755. In a city the size of Denver (pop. 634,265), Romney could only manage less than 70,000 votes?
9 Electoral votes for Obama from a difference of 137,948 votes.
Florida went 50%-49% for Obama. A squeaker. The vote difference was 74,309. Think about all the polling irregularities we heard from Miami-Dade, Broward, and Palm Beach counties, with Republican poll-watchers being kicked out for two hours until the police forced Democrats to let them back in. The vote difference in those three counties alone was 574,033.
29 Electoral votes for Obama from a 74,309 vote difference.
Iowa went 52%-46% for Obama from a 91,927 vote difference. Johnson county had a difference of 26,534 votes, and Linn county had a difference of 20,601 votes. That's half of the state differential right there.
6 Electoral votes for Obama from a 91,927 vote difference.
Virginia went 51%-48% for Obama from a difference of 149,298 votes. Fairfax county had a difference of 87,049 votes. Prince William county had a difference of 28,790 votes. Newport News had a difference of 23,766 votes. Hampton county had a difference of 32,540 votes. These are the heavy blue counties, and they overwhelm the rest of the state.
13 Electoral votes for Obama from 149,298 votes.
Add Ohio's 18 electoral votes from a 166,214 vote difference that came from Cuyahoga county, and it's clear that the election was decided by 1,193,729 from targeted counties in targeted states.
All that is possible in a close election, but not in a true runaway landslide election. I didn't look at 2016 in that kind of detail, but I bet I'd see the same thing. The fact is that this kind of sensitivity to a few key precincts in a few key states suggests that D+10 polls are not reflective of what's really going on.
-PJ