How Were the Presidential Election Polls So Wrong...

[AspieAlly613]

...in 2012?  A lot of discussion has happened about how the 2016 polls (which on average predicted a Clinton 3.5% popular vote victory) ere off by a whopping...1.5%.  Meanwhile, the polls in 2012 suggested that Obama wold win the popular vote by around 1%.  He won by around 4%.

 

The polls were off by twice as much in 2012 as they were in 2016.  Thoughts?

[Guest]

Sounds like people fucking with the system. Things are going to be corrupt no matter what we do. Might as well just blame Russia.

[RoseGoesToYale]

I'd like to know who's doing these dad-blasted polls, because I as a registered voting citizen am never included in them. Who are they polling, paying customers?!

[Cheshire-Cat]

I think the key word is predicted. Generally for the predictions they poll or assume a certain proportion of the population and use that to make a best guess. What actually happens on the day can be totally different depending on whether people bother to vote, and whether they've changed their minds. For example, they may say x part of the country is a sure seat for y and z number of people are likely to vote. Then it turns out maybe they don't vote that way. So in the last election where I am one party had always been very strong, but this time, although they did win, they only won by a very small margin instead of their normal huge margin. And the next time we go to the polls that party may even lose their seat due to the current political climate. Yet in predications they will likely still go with it as a sure seat for that party.

[LeChat]

Well, for those, only a small portion of people are polled. So, it's not 100% indicative of exactly how everyone would vote on election day and what the results will be.

[iff]

Sampling bias is a possible reason but then again, there is also the always present but rarely mentioned, margin of error. We know the polls are not 100%. 

 

A  thing we need to know, is how opinion polling works in that you use previous results to ascertain the  current polling projection. 

 

At the end,I think a little too much emphasis is given to polls without the true understanding of what the number represent as an estimate. 

 

Then there was also the numbers from Rasmussen skewed in romney's favour. Gallup too had a Romney win. 

[Calligraphette_Coe]

59 minutes ago, AspieAlly613 said:

...in 2012?  A lot of discussion has happened about how the 2016 polls (which on average predicted a Clinton 3.5% popular vote victory) ere off by a whopping...1.5%.  Meanwhile, the polls in 2012 suggested that Obama wold win the popular vote by around 1%.  He won by around 4%.

 

The polls were off by twice as much in 2012 as they were in 2016.  Thoughts?

Yeahbut, that's pretty close to the margin of error, usually +/- 3.5 %.

 

I also believe that when the polls are so razor close, it helps turnout. Quite possibly, Obama voters who would have sat out the election if the margins had been a lot bigger instead felt the heat and made sure they voted.

 

I live in one of the states with a lot of electoral college votes, and in the last election, I noticed that the most conservative element waited 3 hours in lines some places. That's what tipped this normally reliable Democratic state just enough ( just a few thousands of votes ) to put it in Trump's column. It was like nothing I had ever seen before! People put yuge signs in their yards cursing Hillary in the meanest ways, and I think so much of that hate went to the polls.

[FindingTheta]

Voter suppression and closing of polling places will certainly tip the balance in favor of Republicans. Plus the more austere you are the more difficult it is to vote.

[Sally]

9 hours ago, Calligraphette_Coe said:

People put yuge signs in their yards cursing Hillary in the meanest ways, and I think so much of that hate went to the polls.

And yet she won the popular vote by almost 3 million votes.  

[Kelly]

On 9/3/2019 at 3:49 PM, AspieAlly613 said:

...in 2012?  A lot of discussion has happened about how the 2016 polls (which on average predicted a Clinton 3.5% popular vote victory) ere off by a whopping...1.5%.  Meanwhile, the polls in 2012 suggested that Obama wold win the popular vote by around 1%.  He won by around 4%.

 

The polls were off by twice as much in 2012 as they were in 2016.  Thoughts?

Which polls? Various polls have differing outcomes. However, the sampling itself has an impact on the accuracy of how its results compare to the entire population of the voters. And it is:

 

 

A large poll (~2400 size) has a margin of error of 2%. The whopping 1.5% that you mention was thus accurate, being within that range.

 

There are other considerations, too. For instance:

 

On 9/3/2019 at 8:47 PM, FindingTheta said:

Voter suppression and closing of polling places will certainly tip the balance in favor of Republicans. Plus the more austere you are the more difficult it is to vote.

Stating for whom you plan to vote and actually voting are not the same thing. If, for instance, you live in a district with a majority of people of color and arrive at your polling station, you might find that it is closed. If it is open, you may need to stand in line for four hours. When you get to the head of the line, you might be denied a ballot or you may have been purged from the voter list.

[AspieAlly613]

6 hours ago, Kelly said:

Which polls? Various polls have differing outcomes. However, the sampling itself has an impact on the accuracy of how its results compare to the entire population of the voters. And it is:

 

 

A large poll (~2400 size) has a margin of error of 2%. The whopping 1.5% that you mention was thus accurate, being within that range.

 

There are other considerations, too. For instance:

 

Stating for whom you plan to vote and actually voting are not the same thing. If, for instance, you live in a district with a majority of people of color and arrive at your polling station, you might find that it is closed. If it is open, you may need to stand in line for four hours. When you get to the head of the line, you might be denied a ballot or you may have been purged from the voter list.

The 1.5% figure was from the 2016 election, but the 2012 election had an error of roughly 3%, twice as much.  The polls conducted had a combined sample of over 14,000 voters.  Using the assumptions that the dominant source of uncertainty is that only a small fraction of the population was polled, one would predict a less than 1/1,000,000,000 chance of an error that great.

 

Of course, you mentioned that there are other considerations too.  These other considerations could cause a systematic, rather than idiosyncratic error, and systematic errors do not decline with sample size.  What's odd is that the systematic error caused the polls to err in favor o the Republicans in 2012, but the Democrats in 2016.