How wrong were the polls in 2016?

[AUFAN78 3,926]

8 minutes ago, homersapien said:

As Ben has already said, we cannot discern what you intended to write beyond what you did write. 

My point is instead of doubling down on your mistake, why not backstep just a little, clarify what you actually intended to mean and correct your wording accordingly? 

Why dispute a literal transcript?  That's irrational.

Trust me, it's not painful at all to correct your wording to reflect what you really meant and apologize for the misunderstanding.  People will actually think more of you for it.

 

You focus on initial statements and your interpretation thereof, but I clarified that early on only to be ignored. It continues.

[homersapien 11,453]

20 minutes ago, AUFAN78 said:

You focus on initial statements and your interpretation thereof, but I clarified that early on only to be ignored. It continues.

No, it looked to me like you refused to back down from your literal words.  You and dub when back and forth several times regarding your literal words.

But I may have missed something.  Please cut and paste it if you think that's the case.

[AUFAN78 3,926]

11 minutes ago, homersapien said:

No, it looked to me like you refused to back down from your literal words.  You and dub when back and forth several times regarding your literal words.

But I may have missed something.  Please cut and paste it if you think that's the case.

I'll give you cred for at least being open to the possibility. Here are a few largely ignored:

"The vast majority of polls missed on this election."

"I can post 100 stories from the days after the election stating the same colossal failure."

"Hundreds of publications in the days following the election stated emphatically how wrong the polls were. I am quite confident they were not talking about margin of error. Again, that is all I tried to say and you lead us down this rabbit trail."

 "Most polls predicted Clinton would be our president. She isn't and they were admittedly wrong."

 

 

[homersapien 11,453]

2 minutes ago, AUFAN78 said:

I'll give you cred for at least being open to the possibility. Here are a few largely ignored:

"The vast majority of polls missed on this election."

"I can post 100 stories from the days after the election stating the same colossal failure."

"Hundreds of publications in the days following the election stated emphatically how wrong the polls were. I am quite confident they were not talking about margin of error. Again, that is all I tried to say and you lead us down this rabbit trail."

 "Most polls predicted Clinton would be our president. She isn't and they were admittedly wrong."

 

 

I think some of the statements above are not factually correct, but more importantly, did you ever acknowledge that what you originally wrote was not literally correct?

[AUFAN78 3,926]

Just now, homersapien said:

But did you ever acknowledge that what you originally wrote was literally incorrect?

I clarified the statement. 

[homersapien 11,453]

Just now, AUFAN78 said:

I clarified the statement. 

You mean you corrected it?

[AUFAN78 3,926]

3 minutes ago, homersapien said:

You mean you corrected it?

No, I clarified it. Polling data was the excuse given by multiple news outlets for getting the prediction wrong. I provided links detailing this fact.

[homersapien 11,453]

7 hours ago, AUFAN78 said:

No, I clarified it. Polling data was the excuse given by multiple news outlets for getting the prediction wrong. I provided links detailing this fact.

Apparently, the hangup here is a technical issue.

If a poll makes a wrong prediction that is nevertheless within the stated margin of error, then the poll is not (technically) wrong.  In other words, in such a close result, either outcome can be supported by that poll's raw data.   

So, your statement that (most of) the polls were "wrong" is simply not true, assuming that most of these polls were within their margin of error, as Dub suggests. (I'll admit that I personally don't know, not having looked into it.)

Now, this may be an esoteric point, but to the scientifically or statistically-trained mind, it's important if evaluating the simple statement "(most) the polls were wrong".   It's not a "trick" question, it's a matter of technical accuracy.  Such accuracy is admittedly more important to some than to others, but it should be considered if making a generalized statement that "the polls were wrong".  You shouldn't blame the polls if the real problem is with people's ability to interpret them.  (Can't solve problems that way.)

The fact that much of the media misunderstands this does not change that reality.

[AUFAN78 3,926]

20 minutes ago, homersapien said:

Apparently, the hangup here is a technical issue.

If a poll makes a wrong prediction that is nevertheless within the stated margin of error, then the poll is not (technically) wrong.  In other words, in such a close result, either outcome can be supported by that poll's raw data.   

So, your statement that (most) of the polls were "wrong" is simply not true, assuming that most of these polls were within their margin of error, as Dub suggests. (I'll admit that I personally don't know, not having looked into it.)

Now this may be an esoteric point, but to the scientifically or statistically-trained mind, it's important if evaluating the simple statement "(most) the polls were wrong".   It's not a "trick" question, it's a matter of technical accuracy.  Such accuracy is admittedly more important to some than to others, but it should be considered if making a generalized statement that "the polls were wrong".  You shouldn't blame the polls if the real problem is with people's ability to interpret them.  (Can't solve problems that way.)

The fact that much of the media misunderstands this does not change that reality.

Some in the media blamed polling. I simply reported it. So perhaps you should take that up with them and I provided links to some of these media outlets to assist with this endeavor. Good luck! 

[homersapien 11,453]

On 1/15/2018 at 4:43 PM, AUFAN78 said:

Some in the media blamed polling. I simply reported it. So perhaps you should take that up with them and I provided links to some of these media outlets to assist with this endeavor. Good luck! 

I don't really care about what the media did or didn't do.

I thought the question on the table is, were the polls generally wrong?  That is a technical question that does interest me.

[AUDub 11,184]

1 hour ago, homersapien said:

Apparently, the hangup here is a technical issue.

If a poll makes a wrong prediction that is nevertheless within the stated margin of error, then the poll is not (technically) wrong.  In other words, in such a close result, either outcome can be supported by that poll's raw data.   

On the eve of the election, the RCP average ended up being Clinton +3.2. The final result was Clinton +2.1. This falls well within the margins of what would be considered a normal polling error. 

Quote

So, your statement that (most) of the polls were "wrong" is simply not true, assuming that most of these polls were within their margin of error, as Dub suggests. (I'll admit that I personally don't know, not having looked into it.)

Of those polls that make up RCP's average:

Bloomberg had Clinton +3. The margin of error is ±3.5. 

IDB had Clinton +1. The margin was ±3.1

The Economist had Clinton +4. They do not provide a MOE. 

LA Times had Trump +3. The margin of error was ±4.8.

ABC had Clinton +3. The margin of error is ±2.5.

Fox had Clinton +4. The margin of error is ± 2.5.

Monmouth had Clinton +6. The margin of error is ±3.5. This is a miss.

NBC had Clinton +5. The margin of error is ±2.73. A miss.

CBS had Clinton +4. The margin of error is ±3.

Reuters had Clinton +5. They do not provide a MOE.

Even if you counted the Reuters and Economist polls among the misses, the majority were still within their margins. 

Quote

Now this may be an esoteric point, but to the scientifically or statistically-trained mind, it's important if evaluating the simple statement "(most) the polls were wrong".   It's not a "trick" question, it's a matter of technical accuracy.  Such accuracy is admittedly more important to some than to others, but it should be considered if making a generalized statement that "the polls were wrong".  You shouldn't blame the polls if the real problem is with people's ability to interpret them.  (Can't solve problems that way.)

The fact that much of the media misunderstands this does not change that reality.

I've said it several times, but this was not a historic polling miss. They were more accurate than they were in the 2012 election. Any who outright says "they were wrong" or asks "where did they miss?" does not really understand how these things work. 

[AUFAN78 3,926]

1 hour ago, homersapien said:

I don't really care about what the media did or didn't do.

I thought the question on the table is, were the polls generally wrong?  That is a technical question that does interest me.

 

[AUFAN78 3,926]

1 hour ago, AUDub said:

On the eve of the election, the RCP average ended up being Clinton +3.2. The final result was Clinton +2.1. This falls well within the margins of what would be considered a normal polling error. 

Of those polls that make up RCP's average:

Bloomberg had Clinton +3. The margin of error is ±3.5. 

IDB had Clinton +1. The margin was ±3.1

The Economist had Clinton +4. They do not provide a MOE. 

LA Times had Trump +3. The margin of error was ±4.8.

ABC had Clinton +3. The margin of error is ±2.5.

Fox had Clinton +4. The margin of error is ± 2.5.

Monmouth had Clinton +6. The margin of error is ±3.5. This is a miss.

NBC had Clinton +5. The margin of error is ±2.73. A miss.

CBS had Clinton +4. The margin of error is ±3.

Reuters had Clinton +5. They do not provide a MOE.

Even if you counted the Reuters and Economist polls among the misses, the majority were still within their margins. 

I've said it several times, but this was not a historic polling miss. They were more accurate than they were in the 2012 election. Any who outright says "they were wrong" or asks "where did they miss?" does not really understand how these things work. 

 

[AUDub 11,184]

2 hours ago, AUFAN78 said:

 

Again, they are doing it wrong. The final popular vote margin ended up not being that far off from the final polling averages. Taken in the aggregate, the RCP average, the miss was 1.2.

The problem, and both you and Bloomberg, as well as many others,  are making the same error, wasn't the polling. It was the assumptions we made from them. 

The track record of polling in American presidential elections is pretty good but a long way from perfect. Obama beat his national polling average by nearly 3 points in 2012. In 2000, Al Gore was behind by about 3 points in the final national polling average but won the popular vote. In 1996, Bill Clinton was ahead in national polls by about 12 points, but won by 8.5. 

You could call the error that occurred routine going by history. 

[homersapien 11,453]

3 hours ago, AUFAN78 said:

 

Sorry, but if that's an argument, I'd prefer you make it in simple text. 

[AUFAN78 3,926]

22 hours ago, AUDub said:

Again, they are doing it wrong. The final popular vote margin ended up not being that far off from the final polling averages. Taken in the aggregate, the RCP average, the miss was 1.2.

The problem, and both you and Bloomberg, as well as many others,  are making the same error, wasn't the polling. It was the assumptions we made from them. 

The track record of polling in American presidential elections is pretty good but a long way from perfect. Obama beat his national polling average by nearly 3 points in 2012. In 2000, Al Gore was behind by about 3 points in the final national polling average but won the popular vote. In 1996, Bill Clinton was ahead in national polls by about 12 points, but won by 8.5. 

You could call the error that occurred routine going by history. 

I think your hang up, one of them anyway, is as Homey mentions a technical one. I think the aforementioned articles/videos hit on technical data, but in the end admit they were wrong due to predicting a Clinton victory. Technically, those are separate issues while related as one prompted the other. 

So to you and perhaps a few others the technical data excuses the pollsters/media even in light of their apologetic commentary. Fine by me. You do you. Hell, you may have a new calling Ben. 

Others however, take these pollsters/media at their word and acknowledge their error. 

 

[AUFAN78 3,926]

19 hours ago, homersapien said:

Sorry, but if that's an argument, I'd prefer you make it in simple text. 

Read response to Ben. Sums it up.

[homersapien 11,453]

2 hours ago, AUFAN78 said:

I think your hang up, one of them anyway, is as Homey mentions a technical one. I think the aforementioned articles/videos hit on technical data, but in the end admit they were wrong due to predicting a Clinton victory. Technically, those are separate issues while related as one prompted the other. 

So to you and perhaps a few others the technical data excuses the pollsters/media even in light of their apologetic commentary. Fine by me. You do you. Hell, you may have a new calling Ben. 

Others however, take these pollsters/media at their word and acknowledge their error. 

 

But I think the underlying - if only implied - question is whether or not polls can be trusted.

[AUDub 11,184]

35 minutes ago, homersapien said:

But I think the underlying - if only implied - question is whether or not polls can be trusted.

And the fact of the matter remains, they can. Anybody shocked by the outcome shouldn't be blaming the national polls, or the nature of polling in general. They performed as well as they ever have. 

[AUDub 11,184]

3 hours ago, AUFAN78 said:

I think your hang up, one of them anyway, is as Homey mentions a technical one. I think the aforementioned articles/videos hit on technical data, but in the end admit they were wrong due to predicting a Clinton victory. Technically, those are separate issues while related as one prompted the other.

They weren't doing it right if they were basing their opinion only on polling. 

Again, I believe Nate Silver nailed it. By his model, it essentially came down to a roll of a three sided die. He caught a lot of hell but was vindicated in the end. 

Quote

So to you and perhaps a few others the technical data excuses the pollsters/media even in light of their apologetic commentary. Fine by me. You do you. Hell, you may have a new calling Ben. 

Others however, take these pollsters/media at their word and acknowledge their error. 

I simply understand how these things work, having a good general understanding of statistics. Certainty is rewarded, but polling is and always will be inexact by its very nature. People need to understand that. If pollsters and forecasters feel the need to apologize, that's fine by me, especially given the opprobrium directed their way by a disappointed public. 

[DKW 86 7,441]

Technical Polling versus Extrapolation versus Influence?

The days before the Election, there was a swarm of talking heads predicting an Electoral Rout by HRC. Maddow, TYT, CNN, most Talking Head shows famously said HRC should win around 320 Electoral Votes. Celebrities said the same thing: Clooney, Hanks, etc.

Here is just an idea. I think the Electoral Rout talkers actually had a suppressive effect on the vote. The 3% was supposed to be spread across the states. Instead, the popular vote totals are actually an overabundance of voter support along the largest coastal states. The margins in the center, especially in the Mid West were much smaller. The talk of an electoral landslide (I think) actually suppressed the vote among HRC Voters. Remember the overwhelming talk just before the vote was landslide. I think just enough voters MAY have stayed home and cost HRC the Election thinking that she has already won the election.

Just my thinking. 

[Grumps 3,704]

I admit to be VERY late to this discussion, but is anyone here saying that the majority of polls were NOT predicting that Mrs. Clinton would win the election?

[AUDub 11,184]

1 minute ago, Grumps said:

I admit to be VERY late to this discussion, but is anyone here saying that the majority of polls were NOT predicting that Mrs. Clinton would win the election?

Nope. Only discussing whether they can be trusted.

RCP average was Clinton +3.2 on election day. Final result was Clinton +2.1. They missed, as all polls do, but they didn't do as badly as people believe. 

[Grumps 3,704]

2 minutes ago, AUDub said:

Nope. Only discussing whether they can be trusted.

RCP average was Clinton +3.2 on election day. Final result was Clinton +2.1. They missed, as all polls do, but they didn't do as badly as people believe. 

Gotcha! From my simple standpoint, the purpose of the polls was to predict the outcome of the election. In that regard it seems that most polls failed. Obviously, in a close election that is more likely to happen, but one would expect, in a close election that some polls would predict one winner, and others would predict another winner. 

[homersapien 11,453]

6 hours ago, Grumps said:

Gotcha! From my simple standpoint, the purpose of the polls was to predict the outcome of the election. In that regard it seems that most polls failed. Obviously, in a close election that is more likely to happen, but one would expect, in a close election that some polls would predict one winner, and others would predict another winner. 

Actually they did, if you only consider the popular vote.

What they didn't do is accurately predict the outcome of the electoral college vote.  As I understand it, that result was by a relatively low number of votes (44k?) distributed in key districts. I'm not sure the precision exists in current methodology to address that scenario.