Were the polls that far off?

Anagram · Nov 9, 2016

One of the major storylines of this election has been the death of public polling. Looking at the results we do have though, that isn't quite so clear. Comparing the RCP average to the current results will show that Arizona, Colorado, Florida, Georgia, Maine, New Hampshire, North Carolina, and Virginia were all within 2% of the poll average. Pennsylvania, Nevada, and National results were within 3%, which is the difference between the polls and national results in 2012. And the national result will probably inch closer to the polling average as the last votes from Seattle and California trickle in.

The big misses were in Ohio (5%), Iowa, (5.5%), Michigan (4%), and Wisconsin (7.5%). The Midwest and Rust Belt went much heavier for Trump than public polls predicted. However, even there the polls were quite accurate with Clinton's percentage while underestimating Trump. And there were reports over the weekend coming out of both sides camps that undecideds in these states were massively going for Trump, and we didn't have any public polls of these states at that point. A plausible explanation is that many people had not fully settled on Trump until the final weekend.

While polling will never have pinpoint accuracy, I don't think it can quite be said that it is dead yet.

Hawkeye10 · Nov 9, 2016

The polls were way wrong....I was completely right to not trust them these last months.

Even the exit polls were way wrong... I made the mistake of trusting them for 1:20 before the election returns straightened me back up.

Anagram · Nov 9, 2016

Hawkeye10 said:
The polls were way wrong....I was completely right to not trust them these last months.

Even the exit polls were way wrong... I made the mistake of trusting them for 1:20 before the election returns straightened me back up.

The exit polls are always wrong. Too much self-selection to get an accurate reading. Why do you think the polls were way wrong though? 7/10 National polls had the margin within 3%. All but one were within 4%.

Fletch · Nov 9, 2016

Anagram said:
One of the major storylines of this election has been the death of public polling. Looking at the results we do have though, that isn't quite so clear. Comparing the RCP average to the current results will show that Arizona, Colorado, Florida, Georgia, Maine, New Hampshire, North Carolina, and Virginia were all within 2% of the poll average. Pennsylvania, Nevada, and National results were within 3%, which is the difference between the polls and national results in 2012. And the national result will probably inch closer to the polling average as the last votes from Seattle and California trickle in.

The big misses were in Ohio (5%), Iowa, (5.5%), Michigan (4%), and Wisconsin (7.5%). The Midwest and Rust Belt went much heavier for Trump than public polls predicted. However, even there the polls were quite accurate with Clinton's percentage while underestimating Trump. And there were reports over the weekend coming out of both sides camps that undecideds in these states were massively going for Trump, and we didn't have any public polls of these states at that point. A plausible explanation is that many people had not fully settled on Trump until the final weekend.

While polling will never have pinpoint accuracy, I don't think it can quite be said that it is dead yet.

None of the polls showed Trump winning 300 electoral votes. Many showed Hillary up by 4, 5, 6 points nation wide when she wasn't. The polls, the pollsters and everyone who used them as a basis for commentary had a fairly easy path to a Clinton victory. Oddly enough, the one hurt most was Hillary, herself. She thought she had it in the bag. The results had to feel like a kick in the crotch to her.

Anagram · Nov 9, 2016

Fletch said:
None of the polls showed Trump winning 300 electoral votes. Many showed Hillary up by 4, 5, 6 points nation wide when she wasn't. The polls, the pollsters and everyone who used them as a basis for commentary had a fairly easy path to a Clinton victory. Oddly enough, the one hurt most was Hillary, herself. She thought she had it in the bag. The results had to feel like a kick in the crotch to her.

None of the polls showed him with 300+ electoral votes, but they did show him very close in enough states for that many. By the end they only showed her up about 3%ish nationally and she's going to win the popular vote by .5-1%ish. They didn't miss by much at all.

And we don't know that Hillary wasn't up by 4, 5, 6 points when the polls showed that. They showed a lot of undecideds during that times, and by all accounts a lot of undecideds came to Trump at the last minute.

Hawkeye10 · Nov 9, 2016

Anagram said:
The exit polls are always wrong. Too much self-selection to get an accurate reading. Why do you think the polls were way wrong though? 7/10 National polls had the margin within 3%. All but one were within 4%.

Because they told a false story on the nature of reality.

Anagram · Nov 9, 2016

Hawkeye10 said:
Because they told a false story on the nature of reality.

Did they? Most of the polls were actually very close. Even in the Midwest states with the biggest margin misses they nailed Clinton's percentage of the vote. They just had tons of undecideds. If most of those undecideds picked Trump at the end, then even those misses weren't actually false at the time.

Hawkeye10 · Nov 9, 2016

Anagram said:
None of the polls showed him with 300+ electoral votes, but they did show him very close in enough states for that many. By the end they only showed her up about 3%ish nationally and she's going to win the popular vote by .5-1%ish. They didn't miss by much at all.

And we don't know that Hillary wasn't up by 4, 5, 6 points when the polls showed that. They showed a lot of undecideds during that times, and by all accounts a lot of undecideds came to Trump at the last minute.

"We hedged enough to claim that we were right even though we told a fantasy story" is not going to work.

Hawkeye10 · Nov 9, 2016

Anagram said:
Did they? Most of the polls were actually very close. Even in the Midwest states with the biggest margin misses they nailed Clinton's percentage of the vote. They just had tons of undecideds. If most of those undecideds picked Trump at the end, then even those misses weren't actually false at the time.

If polls cant predict the winner, and lets be very clear that they did not, then they are useless.

WillyPete · Nov 9, 2016

Anagram said:
One of the major storylines of this election has been the death of public polling. Looking at the results we do have though, that isn't quite so clear. Comparing the RCP average to the current results will show that Arizona, Colorado, Florida, Georgia, Maine, New Hampshire, North Carolina, and Virginia were all within 2% of the poll average. Pennsylvania, Nevada, and National results were within 3%, which is the difference between the polls and national results in 2012. And the national result will probably inch closer to the polling average as the last votes from Seattle and California trickle in.

The big misses were in Ohio (5%), Iowa, (5.5%), Michigan (4%), and Wisconsin (7.5%). The Midwest and Rust Belt went much heavier for Trump than public polls predicted. However, even there the polls were quite accurate with Clinton's percentage while underestimating Trump. And there were reports over the weekend coming out of both sides camps that undecideds in these states were massively going for Trump, and we didn't have any public polls of these states at that point. A plausible explanation is that many people had not fully settled on Trump until the final weekend.

While polling will never have pinpoint accuracy, I don't think it can quite be said that it is dead yet.

It's not so much that the polls are bad, it's that the base assumptions pollsters use to weight the polls didn't apply this time. They've built these over time, and they are largely based on regional/demographic voting trends.

The one poll that apparently was accurate through was the The Los Angeles Times/University of Southern California poll, which has been mocked as a ridiculous outlier, particularly as their readership is very liberal.

As part of their trying to explain the gap, they gave their polling data to other polling groups, who applied THEIR weights to the same data, and came away with a Hillary win, just like all the other polls.

So it's the interpretation of the data that needs work, not the data itself.

Hawkeye10 · Nov 9, 2016

WillyPete said:
It's not so much that the polls are bad, it's that the base assumptions pollsters use to weight the polls didn't apply this time. They've built these over time, and they are largely based on regional/demographic voting trends.

The one poll that apparently was accurate through was the The Los Angeles Times/University of Southern California poll, which has been mocked as a ridiculous outlier, particularly as their readership is very liberal.

As part of their trying to explain the gap, they gave their polling data to other polling groups, who applied THEIR weights to the same data, and came away with a Hillary win, just like all the other polls.

So it's the interpretation of the data that needs work, not the data itself.

Asking the wrong questions of the wrong people is the definition of failing in the social sciences, and that's what happened here. They completely missed the reality of what has happened in this country, as all the Elites did till last night.

Now we see how many are bright enough to catch on, a lot have.

WCH · Nov 9, 2016

Trump never ran a typical campaign. I doubt they will ever figure out what hit them.

Anagram · Nov 9, 2016

Hawkeye10 said:
If polls cant predict the winner, and lets be very clear that they did not, then they are useless.

They predicted the popular vote winner. In the majority of states they predicted the winner and an accurate margin. If there is an abnormally high number of undecideds, how are they supposed to predict that besides reporting there's an unnaturally large number of undecideds, which they did even in the Midwest.

Hawkeye10 · Nov 9, 2016

WCH said:
Trump never ran a typical campaign. I doubt they will ever figure out what hit them.

If they can do their jobs they will, but that is not a given.

THis will take a lot of time though, maybe 3-6 months.

Anagram · Nov 9, 2016

WillyPete said:
It's not so much that the polls are bad, it's that the base assumptions pollsters use to weight the polls didn't apply this time. They've built these over time, and they are largely based on regional/demographic voting trends.

The one poll that apparently was accurate through was the The Los Angeles Times/University of Southern California poll, which has been mocked as a ridiculous outlier, particularly as their readership is very liberal.

As part of their trying to explain the gap, they gave their polling data to other polling groups, who applied THEIR weights to the same data, and came away with a Hillary win, just like all the other polls.

So it's the interpretation of the data that needs work, not the data itself.

That's actually not true. The LA Times poll may end up being the most inaccurate poll of the cycle depending on where Clinton's final margin ends up. They predicted Trump by three when Clinton actually won the popular vote.

Hawkeye10 · Nov 9, 2016

Anagram said:
They predicted the popular vote winner. In the majority of states they predicted the winner and an accurate margin. If there is an abnormally high number of undecideds, how are they supposed to predict that besides reporting there's an unnaturally large number of undecideds, which they did even in the Midwest.

But the polls were wrong about one big thing: They missed Mrs. Clinton’s margin in the Midwestern states, like Wisconsin, Michigan and Pennsylvania.

The exact mechanism for the error is unclear. Perhaps undecided voters broke for Mr. Trump; maybe there really were “silent” voters for him, people who were reluctant to tell pollsters that they backed him. Perhaps it took a lot breaking Mr. Trump’s way: Maybe Republican voters came home to the party over the last week in well-educated suburbs, while undecided white working-class voters broke for Mr. Trump.

But what’s clear is that the error wasn’t simply about the public polls. Even the data team helping Mr. Trump was not forecasting a victory. The Clinton team, which ran its own polls as all campaigns do, was convinced it was on track to victory. It barely even aired advertisements in Wisconsin, Minnesota and Michigan

Why Trump Won: Working-Class Whites - The New York Times

This is where we are today, ya gotta give these people some time to figure out where they went wrong, but they went very wrong, be clear on that point.

Anagram · Nov 9, 2016

Hawkeye10 said:
Why Trump Won: Working-Class Whites - The New York Times

This is where we are today, ya gotta give these people some time to figure out where they went wrong, but they went very wrong, be clear on that point.

They weren't necessarily wrong though.

Example:

Marquette Law poll of Wisconsin on October 31:

Clinton: 46%

Trump: 40%

Johnson: 4%

Undecided/Other: 10%

Actual Results:

Clinton 47%

Trump 48%

Johnson 4%

If that many people really were undecided on October 31, then the poll could be completely accurate. We only know what the final results were. It is entirely possible that those people were undecided last week like the poll said and then broke heavily for Trump on election day.

Mycroft · Nov 9, 2016

Anagram said:
One of the major storylines of this election has been the death of public polling. Looking at the results we do have though, that isn't quite so clear. Comparing the RCP average to the current results will show that Arizona, Colorado, Florida, Georgia, Maine, New Hampshire, North Carolina, and Virginia were all within 2% of the poll average. Pennsylvania, Nevada, and National results were within 3%, which is the difference between the polls and national results in 2012. And the national result will probably inch closer to the polling average as the last votes from Seattle and California trickle in.

The big misses were in Ohio (5%), Iowa, (5.5%), Michigan (4%), and Wisconsin (7.5%). The Midwest and Rust Belt went much heavier for Trump than public polls predicted. However, even there the polls were quite accurate with Clinton's percentage while underestimating Trump. And there were reports over the weekend coming out of both sides camps that undecideds in these states were massively going for Trump, and we didn't have any public polls of these states at that point. A plausible explanation is that many people had not fully settled on Trump until the final weekend.

While polling will never have pinpoint accuracy, I don't think it can quite be said that it is dead yet.

For the most part, the msm-produced polls were as crooked as a three-dollar bill...even up to the end.

After Trump won his Party's nomination, the polls were constantly manipulated to advance the narrative that Hillary had him beat. They constantly injected outrageous +12, +10, +8, etc poll results to keep Hillary's RCP average up. This served to make her look like she was doing better than she was. Even toward the end...when polls generally even out and become accurate...those guys couldn't help but cook the numbers to keep making Hillary look good.

The media and their paid pollsters will reap the rewards of their actions. It is unlikely the majority of Americans will ever believe them again.

AlbqOwl · Nov 9, 2016

Anagram said:
One of the major storylines of this election has been the death of public polling. Looking at the results we do have though, that isn't quite so clear. Comparing the RCP average to the current results will show that Arizona, Colorado, Florida, Georgia, Maine, New Hampshire, North Carolina, and Virginia were all within 2% of the poll average. Pennsylvania, Nevada, and National results were within 3%, which is the difference between the polls and national results in 2012. And the national result will probably inch closer to the polling average as the last votes from Seattle and California trickle in.

The big misses were in Ohio (5%), Iowa, (5.5%), Michigan (4%), and Wisconsin (7.5%). The Midwest and Rust Belt went much heavier for Trump than public polls predicted. However, even there the polls were quite accurate with Clinton's percentage while underestimating Trump. And there were reports over the weekend coming out of both sides camps that undecideds in these states were massively going for Trump, and we didn't have any public polls of these states at that point. A plausible explanation is that many people had not fully settled on Trump until the final weekend.

While polling will never have pinpoint accuracy, I don't think it can quite be said that it is dead yet.

The morning of the election, the betting odds were 85% Clinton to 15% Trump and that was based on the RCP polling average.

But most of the RCP polling average was based on polls by the MSM and groups who are more partisan than not and I honestly think that did skew the results they produced toward Hillary. It is no secret that Hillary was the preference of most of those doing the polling.

The other thing is the backlash against what was perceived as stacking the deck in Hillary's favor. I know so many people who were giving quite dishonest responses to the pollsters when they called. And there was an awful lot of polling done in the last two or three weeks, enough to produce burnout on the part of those receiving those calls. Why were they dishonest? Right or wrong, it was just a way to screw with polling perceived to be self serving instead of an honest effort to produce an honest result.

Anagram · Nov 9, 2016

Mycroft said:
For the most part, the msm-produced polls were as crooked as a three-dollar bill...even up to the end.

After Trump won his Party's nomination, the polls were constantly manipulated to advance the narrative that Hillary had him beat. They constantly injected outrageous +12, +10, +8, etc poll results to keep Hillary's RCP average up. This served to make her look like she was doing better than she was. Even toward the end...when polls generally even out and become accurate...those guys couldn't help but cook the numbers to keep making Hillary look good.

The media and their paid pollsters will reap the rewards of their actions. It is unlikely the majority of Americans will ever believe them again.

That is an interpretation of what the polls did, but I don't believe it's the correct one. Although there were a few polls showing results like +8 and +12 her polling average never was above 7.1% and that was simply two short lived bounces after the Democratic convention and after the Access Hollywood tape. More importantly though, Hillary's average was never above 45% in the polls. (She's almost at 48% now). An extremely plausible interpretation of the polls is that for most of the campaign there really were many undecided voters and that at the end they heavily broke for Trump. This intuitively makes sense. Trump got many, many working class voters who had been voting Democratic their whole life. It makes sense they didn't make up their mind to change their voting philosophy until right before the election.

AlbqOwl · Nov 9, 2016

AlbqOwl said:
The morning of the election, the betting odds were 85% Clinton to 15% Trump and that was based on the RCP polling average.

But most of the RCP polling average was based on polls by the MSM and groups who are more partisan than not and I honestly think that did skew the results they produced toward Hillary. It is no secret that Hillary was the preference of most of those doing the polling.

The other thing is the backlash against what was perceived as stacking the deck in Hillary's favor. I know so many people who were giving quite dishonest responses to the pollsters when they called. And there was an awful lot of polling done in the last two or three weeks, enough to produce burnout on the part of those receiving those calls. Why were they dishonest? Right or wrong, it was just a way to screw with polling perceived to be self serving instead of an honest effort to produce an honest result.

As a mea culpa re my post quoted here, it is interesting reviewing the fivethirtyeight group's analysis that the MSM seems to have done considerably better in this election cycle than many of the private polling groups:

http://projects.fivethirtyeight.com/pollster-ratings/

Anagram · Nov 9, 2016

AlbqOwl said:
The morning of the election, the betting odds were 85% Clinton to 15% Trump and that was based on the RCP polling average.

But most of the RCP polling average was based on polls by the MSM and groups who are more partisan than not and I honestly think that did skew the results they produced toward Hillary. It is no secret that Hillary was the preference of most of those doing the polling.

And those polls weren't that far off, ending at Clinton +3%. When California and Washington finish their results it will probably be around a 1% Clinton win. Hardly a catastrophic miss.

And the areas where the margins were off, Pennsylvania, Ohio, Wisconsin, etc., there's a plausible interpretation of why that is. They didn't overestimate Clinton's eventual vote percentage. They underestimated Trump while having large numbers of undecideds. If there really were that many undecideds in the last week, those who voted Democratic their whole life but might finally pull the lever for Trump for instance, the polls in those states weren't even incorrect at the time. It's just a group that went heavily for Trump hadn't made up their minds.

calamity · Nov 9, 2016

Anagram said:
One of the major storylines of this election has been the death of public polling. Looking at the results we do have though, that isn't quite so clear. Comparing the RCP average to the current results will show that Arizona, Colorado, Florida, Georgia, Maine, New Hampshire, North Carolina, and Virginia were all within 2% of the poll average. Pennsylvania, Nevada, and National results were within 3%, which is the difference between the polls and national results in 2012. And the national result will probably inch closer to the polling average as the last votes from Seattle and California trickle in.

The big misses were in Ohio (5%), Iowa, (5.5%), Michigan (4%), and Wisconsin (7.5%). The Midwest and Rust Belt went much heavier for Trump than public polls predicted. However, even there the polls were quite accurate with Clinton's percentage while underestimating Trump. And there were reports over the weekend coming out of both sides camps that undecideds in these states were massively going for Trump, and we didn't have any public polls of these states at that point. A plausible explanation is that many people had not fully settled on Trump until the final weekend.

While polling will never have pinpoint accuracy, I don't think it can quite be said that it is dead yet.

Biggest problem with polling was poor turnout. Hillary pulled in 6 million fewer votes than Obama, and in states like PA, MI, WI, NC, and FL that was a killer. She lost some of those states by margins less than the number of of black voters who voted for Obama but stayed home yesterday: places like Detroit, Philly, Milwaukee, Charlotte, and Miami-Dade.

TaterTots · Nov 9, 2016

With the way people assume Trump is a racist and his supporters are, I'd be hard pressed to admit I was voting for him too. If you state you voted for him you'd have 12 people calling you a racist because the left wing media has brainwashed them. Not worth the hassle.

Mycroft · Nov 9, 2016

calamity said:
Biggest problem with polling was poor turnout. Hillary pulled in 6 million fewer votes than Obama, and in states like PA, MI, WI, NC, and FL that was a killer. She lost some of those states by margins less than the number of of black voters who voted for Obama but stayed home yesterday: places like Detroit, Philly, Milwaukee, Charlotte, and Miami-Dade.

If the polls had been accurate from the get go, they would have predicted her poor turnout.

Were the polls that far off?

Anagram

Hawkeye10

Buttermilk Man

Anagram

Fletch

Anagram

Hawkeye10

Buttermilk Man

Anagram

Hawkeye10

Buttermilk Man

Hawkeye10

Buttermilk Man

WillyPete

The sound of one chao mooing.

Hawkeye10

Buttermilk Man

WCH

Believer

Anagram

Hawkeye10

Buttermilk Man

Anagram

Hawkeye10

Buttermilk Man

Anagram

Mycroft

Genius is where you find it.

AlbqOwl

Anagram

AlbqOwl

Anagram

calamity

Privileged

TaterTots

Active member

Mycroft

Genius is where you find it.