How did the pollsters get it wrong AGAIN?

[Torn Mind]

Maybe in a deep blue state, but I doubt that Ohio or Florida voters are reluctant to admit to a pollster that they're voting for him.

Florida is not a 60/40 state. It's a 50/50 state. Ohio is still closer to 50 percent than 60.

Florida is also very big, very populous, and has crappy summer weather to kill its property values. So it attracts people from the outside looking to live on the cheap, including those who grew up blue.

 

[TeeJay1952]

I agreed with Hillary that there was a 10 to 15% basket of deplorables.
Trump showed me the basket is MUCH larger.
No matter what happens I am devastated.

 

[Ajay]

Pollsters should be dragged out and shot after this debacle. How do they make any money after being to monstrously incorrect?
I'm sure they'll all have their normal excuses as to why the data they collected was misleading.

 

[Indus]

Maybe they didn't get it wrong but the mail in ballots are *lost* in the mail?

 

[fskimospy]

I suspect the polls will look considerably less wrong in about 2-3 weeks.

 

[TheVrolok]

Maybe you have missed the terrorist organization called BLM. They target houses with Trump signs.

I'm going to enjoy this Trump win more so than the last. I can't wait to see the riots from the snowflakes and the celebrities crying on TV. Its going to be fucking glorious.

Hey @frowertr, enjoying all that glory?

It's ok, we know you won't be coming back.

 

[zinfamous]

Reality is what we perceive it to be. Unfortunately, millions of Americans live in a reality vastly different than ours. A reality where Democrats eat babies and always have.

but they're so delicious!

 

[zinfamous]

The election isn't over and it will take a while to unpack everything but there seems to be general confusion about how campaign internals on both sides and district level polling didn't even hint much at this outcome. The Rs seem just as surprised as the Ds. Overall a very strange election.

I think what happened is that Biden got a lot of GOP votes, but those voters still voted R down ticket. Especially in those states where Dems thought they were gaining ground in Senate races--Maine, NC, GA, etc.

If I had to guess, polsters were interpreting GOP votes for Trump as also GOP rejection of the whole of the GOP.

The presidential results seem completely normal from the various predictions. This was the second-to-least bad option that people were expecting. And, because it wasn't an obvious bloodbath from the beginning, it meant the senate and house races were a lot closer than people thought.

 

[zinfamous]

IDK Is it the echo chamber effect? We block those we disagree with and that screws up perceptions?
I tip my hat to my conservative brothers and sisters and wonder how we will explain this to history.
I am the saddest MFer in history but I will survive.

wow, really? Where you born in Aushwitz or something?

 

[zinfamous]

If the polls are off within the margin of error and some are skewed toward Biden and some are skewed toward Trump, that's one thing, but that's not what we're seeing. This is the second GE in a row where there the polling was off pretty consistently in one direction.

Only theory that I've seen so far that makes sense to me is that there is a certain segment of Trump's base that refuses to answer any call from unknown numbers. Obviously a lot of people do this, myself included, but I guess the theory is that this particular segment is less likely to answer these calls than whatever the polling base is.

I know I don't answer unknown calls. I also get all my calls from CA, because my # is still CA, while I live in MD, so it's easy for me to assume they are garbage calls. All the text spam that I get it is local CA elections and proposition begging. I get very few MD calls, but I still only answer maybe half of them, when I am expected a call from a bank or clinic or, whatever. Still, all of my MD residency forms that I have signed through the years--DMV, mortgage, employment, whatever, have the same phone number....so I'd expect that I still get MD calls for campaigns, no?

Or do campaigns not use state registration data to make their calls? Do they just go through lists of registered phone numbers? With cell phones, area codes no longer mean anything. And democrats, I'd think, are more likely to move around the country. Where I've worked over the years, my lab groups are usually constructed of people from all over the country, with various area code phone numbers and of course many of them are just international.

I honestly can't recall the last time that I've been polled....but now that I think about it, living in Chicago, Berkeley, DC area over the last 15 years might explain it. No one is interested in those voting trends.

 

[K1052]

I think what happened is that Biden got a lot of GOP votes, but those voters still voted R down ticket. Especially in those states where Dems thought they were gaining ground in Senate races--Maine, NC, GA, etc.

If I had to guess, polsters were interpreting GOP votes for Trump as also GOP rejection of the whole of the GOP.

The presidential results seem completely normal from the various predictions. This was the second-to-least bad option that people were expecting. And, because it wasn't an obvious bloodbath from the beginning, it meant the senate and house races were a lot closer than people thought.

Though generic ballot question was a big miss too for most pollsters looks like.

I know the GOP is happy, relatively, with the result but they're also worried that they can't trust their numbers either going forward.

 

[fskimospy]

Seems relevant to note that the pollsters did predict Biden would win, did predict that the line between a blowout and a thin victory was close, and did predict that Pennsylvania was the likely deciding state.

 

[Blackjack200]

I know I don't answer unknown calls. I also get all my calls from CA, because my # is still CA, while I live in MD, so it's easy for me to assume they are garbage calls. All the text spam that I get it is local CA elections and proposition begging. I get very few MD calls, but I still only answer maybe half of them, when I am expected a call from a bank or clinic or, whatever. Still, all of my MD residency forms that I have signed through the years--DMV, mortgage, employment, whatever, have the same phone number....so I'd expect that I still get MD calls for campaigns, no?

Or do campaigns not use state registration data to make their calls? Do they just go through lists of registered phone numbers? With cell phones, area codes no longer mean anything. And democrats, I'd think, are more likely to move around the country. Where I've worked over the years, my lab groups are usually constructed of people from all over the country, with various area code phone numbers and of course many of them are just international.

I honestly can't recall the last time that I've been polled....but now that I think about it, living in Chicago, Berkeley, DC area over the last 15 years might explain it. No one is interested in those voting trends.

I was listening to a podcast yesterday and they made this exact point. These polls can no longer be considered representative polls of voters, they are polls of people that answer their phones for random phone numbers and then are willing to talk to pollsters. I don't know what they do next cycle, because this is just not reliable anymore.

 

[Ken g6]

Seems relevant to note that the pollsters did predict Biden would win, did predict that the line between a blowout and a thin victory was close, and did predict that Pennsylvania was the likely deciding state.

On the other hand, pollsters were off about many of the congressional races. Instead of a "blue wave", it seems to me we got more of a "red tide".

 

[fskimospy]

On the other hand, pollsters were off about many of the congressional races. Instead of a "blue wave", it seems to me we got more of a "red tide".

I think the most accurate description is a blue wave combined with a somewhat smaller red wave. You are right though that the overall congressional count was a pretty big miss though.

 

[pmv]

I was listening to a podcast yesterday and they made this exact point. These polls can no longer be considered representative polls of voters, they are polls of people that answer their phones for random phone numbers and then are willing to talk to pollsters. I don't know what they do next cycle, because this is just not reliable anymore.

I think the second part there is not new - they've always had to allow for the fact that many won't be prepared to talk to pollsters. But I wonder if the first bit, about not answering the phone at all for unknown numbers, may have gotten worse over time? Due to the growth in telephonic 'spam', plus more widespread use of 'caller ID'. People are probably more wary of time-wasting marketing and scam calls now than they used to be.

 

[Ajay]

If I had to guess, polsters were interpreting GOP votes for Trump as also GOP rejection of the whole of the GOP.

Maybe I have the wrong reading glasses on, but what are you trying to say?

 

[IronWing]

Maybe I have the wrong reading glasses on, but what are you trying to say?

You have to remember that Zin's been trying to keep up with Someone, beer for beer.

Putting on the beer goggles reveals:
If I had to guess, pollsters were interpreting GOP votes against Trump as also GOP rejection of the whole of the GOP.

 

[dullard]

Pollsters should be dragged out and shot after this debacle. How do they make any money after being to monstrously incorrect?
I'm sure they'll all have their normal excuses as to why the data they collected was misleading.

The thing is, the polls look to be as accurate as one should expect this time around.

1) Two swing states (North Carolina and Florida) were miss-called but they were always listed as very close. And they were very close. Looks like the other 48 states were correctly called by the polls.

2) Once all mail-in ballots are counted, it looks like it will be Biden over Trump by 4 to 5%. The polls averaged Biden over Trump by 8% (+-6%). So, the final result is within the polls margin of error.

The part that was monstrously incorrect was people thinking that polls are more accurate than they are. Polls are usually +-3% PER CANDIDATE. Thus, when subtracting the two numbers, you need to include the error for both candidates. This makes a Biden>Trump or Trump>Biden prediction have a +-6% margin of error. I'm slightly oversimplifying the math here, to avoid a long statistics discussion, but doubling the per candidate error when the candidates are close gives a close-enough result.

Suppose the final count ends up being 52% Biden to 48% Trump (ignoring the "none of the above" type results), then you are within 538's final predictions from the polls (see the light blue and light pink 80% of outcome bands below).

 

[Ajay]

The thing is, the polls look to be as accurate as one should expect this time around.

1) Two swing states (North Carolina and Florida) were miss-called but they were always listed as very close. And they were very close. Looks like the other 48 states were correctly called by the polls.

2) Once all mail-in ballots are counted, it looks like it will be Biden over Trump by 4 to 5%. The polls averaged Biden over Trump by 8% (+-6%). So, the final result is within the polls margin of error.

The part that was monstrously incorrect was people thinking that polls are more accurate than they are. Polls are usually +-3% PER CANDIDATE. Thus, when subtracting the two numbers, you need to include the error for both candidates. This makes a Biden>Trump or Trump>Biden prediction have a +-6% margin of error. I'm slightly oversimplifying the math here, to avoid a long statistics discussion, but doubling the per candidate error when the candidates are close gives a close-enough result.

Good points. Thanks. I haven't dealt with statistical and systematic errors in, uh, 25+ years - sucks getting older, some old stuff fades back into the noise.

 

[vi edit]

The part that was monstrously incorrect was people thinking that polls are more accurate than they are. Polls are usually +-3% PER CANDIDATE. Thus, when subtracting the two numbers, you need to include the error for both candidates. This makes a Biden>Trump or Trump>Biden prediction have a +-6% margin of error. I'm slightly oversimplifying the math here, to avoid a long statistics discussion, but doubling the per candidate error when the candidates are close gives a close-enough result.

Yeah I think this is the real key here. I think a lot of us tend to focus on *just* one of those sides of the margin. "Oh my guy is up by 5, cool!" And they just assume the worse for the other candidate and think it's safe.

The reality is if Biden is at 52 and Trump is at 48 with a +/- 3%. That could actually be a Trump win at 51/49.

If there's any real shock and introspection to this election, it's where and how did Trump get that many more voters to turn out. I'm not saying their was anything shady, not at all. Just simply they got like 5+ million more voters out this year than in 2016. I'd be very interested to see where those votes came from and how the engagement to get them out to vote happened.

Say what you want about Trump and his personal campaign approach, their ability to drum up votes over 2016 was impressive.

 

[dullard]

If there's any real shock and introspection to this election, it's where and how did Trump get that many more voters to turn out. I'm not saying their was anything shady, not at all. Just simply they got like 5+ million more voters out this year than in 2016. I'd be very interested to see where those votes came from and how the engagement to get them out to vote happened.

Say what you want about Trump and his personal campaign approach, their ability to drum up votes over 2016 was impressive.

Trump was correct about a large "silent majority". We've always had millions and millions of registered voters (or people who could become registered) that just don't vote. What Trump was wrong about was that all of that silent majority wants him to win. Trump did one thing very well: he excited his base. That also however, angered millions others enough to bother to vote against him too.

 

[TheVrolok]

Maybe you have missed the terrorist organization called BLM. They target houses with Trump signs.

I'm going to enjoy this Trump win more so than the last. I can't wait to see the riots from the snowflakes and the celebrities crying on TV. Its going to be fucking glorious.

Figured I'd check in again, @frowertr, how's the glory feeling?