On the blind squirrel finding a nut in 2016

[JohnStOnge]

I'm talking about the Rasmussen poll. If you look at 538's assessment of its performance over a large number of polls it's clearly not one of the better polls. Of 396 polls on the 538 list Rasmussen ranks 199th. Middle of the pack. If you look just at the 46 polls such that 538 had at least 30 polling results to review, it's ranked 30th. It is tied for being he 19h most biased poll, historically, among the total 396 polls evaluated and tied for being the 8th most biased among the 46 with at least 30 results to review.

Nevertheless, people assign it credibility because it's point estimate for the 2016 election was Clinton by 2 and Clinton won by 2.1. The blind squirrel found a nut.

But what happened this time? Rasmussen's point estimate, Republicans by 1, was THE farthest off on the overall popular vote for the 2018 House elections among the 11 poll point estimates to calculate the latest RealClearPolitics (RCP) average (https://www.realclearpolitics.com/epoll ... -6185.html). Right now the latest available vote count gives the Democrats a 6.5 percentage point edge. That means the Rasmussen point estimate was off by around 7.5. The next most off point estimate is CNN's at 6.5.

Also, Rasmussen is the only one of the polls such that it really was "wrong" in that the actual result is outside of its margin of error. Considering the reported margin of error means the Rasmussen poll said it was between Democrats by 1 and Republicans by 3. CNN, the poll with the closest call to "missing" among the other polls, reported a margin of error that means it was between Democrats by 6 and Democrats by 20.

Fellas, the Rasmussen poll is viewed by most people who follow polling as relatively unreliable for a reason. Actually a number of reasons including its methodology and it's actual historical performance.

[JohnStOnge]

Since my first post was so long and I know you guys don't like that I'll at least try to REDUCE the length of a single post by including a comparison I did in this one. I worked my way back using the RealClearPolitics (RCP) pages on various national elections I could find to identify all the national election situations I could whereby both the ABC News/Washington Post poll and the Rasmussen Poll were part of the final RCP average. I ended up with 2018 House Elections, 2016 Presidential Election, 2014 House Elections, 2012 Presidential Election, 2010 House Elections, and 2008 Presidential Election.

For those races the Rasmussen poll has more than twice the average simple error, at 3.6, than the ABC News/Washington Post poll does at 1.6. It also has more than twice the bias, at 2.7 in favor of Republicans, as the ABC News/Washington Post poll does at 1.2 in favor of the Democrats.

The ABC News/Washington Post poll point estimate fell on the side of the side that won all six times. The Rasmussen poll point estimate fell on the wrong side two times in only six tries as it had Romney winning the popular vote in 2012 and the Republicans winning the popular vote this year.

The ABC News/Washington Post poll's point estimate maximum error was 2.8 percentage points. It had Republicans by 4 for the 2010 House elections and they won by 6.8. The Rasmussen poll point estimate missed by about 7.1 points for this year's House elections, missed by 4.9 points when it had Romney beating Obama in the popular vote, and missed by 5.2 points when it had the Republicans by 12 in the overall popular vote and they won it by 6.8. As you can see all three of those large misses missed in the direction of over estimating Republican performance.

I picked the ABC News/Washington Post poll for comparison because, as you may recall, it is rated A+ by 538. As you also may recall the Rasmussen poll is rated C+.

Note: The numbers for the 2018 House election will change but all indications are the Democrat margin is likely to increase. If that's the case the Rasmussen poll will look even worse.

[mainejeff]

Too many times we use soundbites to explain modern day politics. I appreciate your research and unbiased approach.

[JohnStOnge]

I can't confirm because as far as I could tell I'd have to pay to look at the Rasmussen poll cross tabs. But I suspect Trump is referring to Rasmussen poll results when he says he has high approval among Blacks.

He clearly does not. The two most recent polls up at his job approval page at RCP, Economist/YouGov and NPR/PBS/Maris, both have him at 10% job approval among Blacks. And, while it's not a direct vote on Trump, the exit polling estimate that Democrats beat Republicans by 90% to 9% among Blacks in House races is consistent with that general level of support.

[Skjellyfetti]

Thought you'd like these charts, JSO

https://www.nytimes.com/interactive/201 ... lysis.html

[JohnStOnge]

Thought you'd like these charts, JSO

https://www.nytimes.com/interactive/201 ... lysis.html

That's interesting. The charts represent House elections exit polls and are there are things that are different than Presidential election polls. You can see the historical Presidential election poll results at https://ropercenter.cornell.edu/polls/u ... ups-voted/.

18 to 29 year olds voted for Reagan by an estimated 59% to 41% in 1984 so that line would've gone way over to near the +20 line on the red side then. No indication of anything like that, apparently, in the House exit polls.

But I think the general trends are consistent. Younger age groups have become more likely to vote Democrat while older age groups have become more likely to vote Republican over time.

One trend they kind of mentioned involves Asians. When they first started tracking Asians in Presidential election exit polling in 1992 they voted by an estimated 55% to 31% Republican. In 1996 it was 48% to 44% Republican. But it's been on the plus side for Democrats ever since. A clear downward trend in Republican support. If you list the %Republican support for each election 1992 - 2016 in Presidential elections it's 55%, 44%, 41%, 43%, 35%, 26%, and 27%. You can see there's a strong trend towards Asians liking the Republican candidate less as time went on. And I think the graphic at the page you linked also shows such a trend. Not as strong, but still clearly one in which Asians have been moving away from supporting Republicans over time. This last time only 23% of Asians voted Republican.

Yet the chart shows Asians moving into Republican territory during the 2014 mid terms. I wonder what hacked them off that year. You can see that they were well into the blue during the 2012 and 2016 House elections. And the Democratic Presidential candidates won among Asians by an estimated 73% to 26% in 2012 then 65% to 27% in 2016. Weird that they'd swing over like that for one year. In any case they were farther into the blue than ever this year.

[Ivytalk]

Thought you'd like these charts, JSO

https://www.nytimes.com/interactive/201 ... lysis.html

Fvcking enabler!

[Baldy]

Thought you'd like these charts, JSO

https://www.nytimes.com/interactive/201 ... lysis.html

Fvcking enabler!

JSO being the pivot man in a gay circle jerk...he's in heaven.

[JohnStOnge]

Fvcking enabler!

JSO being the pivot man in a gay circle jerk...he's in heaven.

Just doing my public duty by letting you guys know that the fact that the Rasmussen poll had on instance where its point estimate was closest does not mean it is a relatively reliable poll. It's not.

[CID1990]

What is this thread about?

[Ivytalk]

What is this thread about?

Rasmussen’s nuts.

[Winterborn]

What is this thread about?

Rasmussen’s nuts.

What type of squirrel is a Rasmussen? And more importantly, do they taste good?

[andy7171]

What is this thread about?

JSO blathering on how to prove Trump is worse than Hillary.


Sent from my iPhone using Tapatalk

[JohnStOnge]

What is this thread about?

JSO blathering on how to prove Trump is worse than Hillary.


Sent from my iPhone using Tapatalk

No no no. JSO helping you guys out by demonstrating that the Rasmussen poll is not reliable so that those of you who have cited it in the past can avoid doing that in the future.