« October 23rd 1983, Beirut Bombings | Main | The Turks want Mosul and Aleppo "back." »

23 October 2016

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

Fifth Columnist

It's called an outlier. Instead of over-valuing a single poll you should be looking at the trends for poll averages.

This poll shows Clinton winning ~77% of Democrats and only 65% of blacks. Not likely to be accurate.

In any case it makes no difference. Trump is certainly campaigning as if he's down big with all the rigged election whining.

turcopolier

5th Colummist

Rasmussen is another outlier? I suspect that there is a lot of group think in the polling world as well as commonly shared sampling biases. pl

Jack

Sir

Zero hedge has two recent posts raising questions on the poll results.

One about the skews.

http://www.zerohedge.com/news/2016-10-22/statistician-warns-americans-ignore-capricious-polls

Second about how WaPo came up with the Borg Queen up 12 points in their latest poll.

http://www.zerohedge.com/news/2016-10-23/how-wapos-latest-poll-give-hillary-12-point-advantage-over-trump

Now I'm no statistician and really don't have any particular insight on the mechanics of polls. What I'm befuddled by is the emphasis on national polls when the presidential election is done on a state basis. Are the state polls equally flawed with their sampling? And what are their models for turnout?

What I've noticed as you have pointed out is the massive effort by the Borg amplified by their media to destroy Trump to ensure the Borg Queen's crowning, yet Trump within the margin of error in the polls in most battleground states. What's up with that? Is there more latent support for Trump or more Les Deplorables than seems evident in the MSM election prognosis?


David Habakkuk

All,

My wife’s hairdresser is an excellent source for local gossip. After the Brexit result, he told her that a well-known BBC presenter had been in to have his hair done on polling day, and been confident that ‘Remain’ would win by 10%. And the ‘trends for poll averages’ turned out wrong.

LondonBob

Richard Baris says IDB/TIPP has it as a D+8 as well. He sees it as being a D+4 to +6 election and also has Trump two points ahead. Maybe this is all they have left, claim the race is over and hope to suppress the Trump vote? Always watch where the campaigns are, same battleground states, except looks like the Trump campaign sees Iowa as in the bag and isn't bothering there anymore.

Times like this I wish I hadn't forsworn gambling, Trump at 4-1 at the bookies.

Fifth Columnist

Sure, any poll that doesn't track with the polling averages could be considered an outlier. That doesn't mean that the IBD or Rasmussen polls are provably wrong but that their results should perhaps be considered more unlikely than not.

If future polling or the election on Nov 8th validates these current polls then we can revisit their merit.

I say it doesn't matter because these polls should have no practical effect on voting. Non-pollsters and those who have no background in statistical methodology have no way to determine what polls are credible. That's why I suggest looking at poll averages.

I think political observers find this confusing when their personal indicators, such as numbers of lawn signs or crowd sizes at rallies, don't match the media polls.

Jay

I am not putting much trust in the polls this time. Trump has no common frame of reference when it comes to politics. The polls like himself. Are All over the place.

Jay

I don't know about the "rigged election whining". Between, the press, Wikileaks and what just happened in Illinois I'd say he has made some serious points.

turcopolier

5th columnist

In all the years leading to my present advanced decrepitude one thing I learned is that the massed opinions of incorrect analysts grouped together for mutual support and security are apt to be wrong while the outliers in the back of the room are often right. I am interested in who is right not the statistical crap or your logrolling. I still think she will win but I think it will be close in the popular vote and in the Congress. pl

LondonBob

It isn't an outlier, most reliable polls (LAT, PPD as well) have Trump narrowly ahead.

johnf

Our polls were not only wrong in the Euro referendum but in last year's General Elction. Only one or two outliers were correct.

Dan Berg

these polls are done over the telephone; the only people who still have land lines are the elderly, therefore................

johnf

So convinced was I by the polls that I didn't even bother to stay up and watch. I had a pleasant awakening.

Dan Berg

polls are done over the telephone; the only people who still have land lines are the elderly, therefore. . . . .

Tyler

Sir,

Oversampling Dems.

Questions designed to weed out Repubs and right leaning independents.

Ignoring Trump's base (poor whites who didn't vote for Romney)

Case in point there was a "poll" in AZ claiming HRC was up by 5. It was screamed from the hillyops as proof of a landslide. Problem is that it had a whopping D+38 oversample. This is all psyops onTeump voters trying to convince them it's hopeless and to not vote. More media collusion that Wikileaks and Okeefe proved.

I can't even imagine what the internal polls look like for HRC.

Tyler

LB,

I put $500 on Trump

Eric newhill

Sir, I have concluded that the most of the polls are some of that "we create our own reality stuff".

Tactically, it seems risky to artificially inflate the polls in favor of Clinton because it could cause her weaker supporters to decide she's got it and just stay home on voting day. However, there is probably some committee of genius pollster operatives that decided that the demoralization of Trump support would outweigh the complacency suppression on the Clinton side. These pollsters are either going to be heroes or schmucks of the first order on Nov 9.

JohnH

Having contracted for market research surveys for my company, I have become more suspicious of survey results over the years.

First, any poll is a survey not of the population but only of those willing to respond. Internet surveys are particularly bad in this regard, because they only include those motivated to take the initiative to respond.

Second, my impression is that the number of those willing to respond has declined significantly over the years. As cell phones became more common, pollsters could reach a smaller and smaller part of the population, because cell phone numbers are not published. Also, as caller ID became more common, people had the option to answer calls from unknown numbers or to not answer. As phone telemarketing scams became more and more common, often under the guise of research, more and more people decided not to respond to any call identified as 'research.'

Personally, I won't answer any call from an unrecognized or identified source. IMO those who do answer such calls tend to be populations with low tech land lines...older and poorer.

Fredw

The breakdown on this poll indicates that Trump's <2 point lead consists entirely of an 8-point lead in the south. He's behind in all other regions. (By about 2 percent in each.) He may be suffering from a reverse of the Democrats' congressional imbalance, in which they have more votes but all concentrated in a few districts.

Jack

LondonBob

The bookies always make money like the house in a casino. The odds they give is based on their book. I've read there is more money on the Borg Queen winning but more bets albeit with less money on Trump, similar to the Brexit bets in your country. I too will take the odds on Trump.

Jack

All

And more Hillarity.....

http://www.zerohedge.com/news/2016-10-22/heres-30-seconds-after-last-debate-cnn-would-rather-you-didnt-see

Checkout who were undecided in the front row!

kao_hsien_chih

People are mixing up (inherent) uncertainty in public opinion and margins of error (that is the problem that the article Jack linked to was getting at.) Even if pollsters are measuring the same thing, from samples taken from the same population, they can be off by a little bit because of sampling errors--that's margin of errors. When different polls are farther from each other than by the margins of error (or twice that), the short answer is that they are measuring something different (In fact, even if they are within margins of error, if they are sufficiently far apart, they probably are measuring something different, just that the odds are less than the usual statistical convention.)

The trouble with polls is that since public opinion data is limited and nobody knows exactly what will happen on the election day, a lot of guesses go into forming the prediction, especially about what the electorate will look like on the election day. In 2012, most pollsters had similar ideas about who will show up, which, it turned out, was justified by the actual events. Most polls predicted something similar so that adding up the poll numbers gave clearer idea about what the election results would be than any single poll. In 2016, pollsters are predicting wildly different results basically because they have very different ideas about what the electorate will look like, into which they are fitting the public opinion data. Simply adding up the polls doesn't help much because they are not even talking about something close to one another. The bottom line is that we literally don't know as much as we'd like, and we are messing with numbers and formulas that don't fit without having a clear idea what we are talking about (perhaps not quite that bad.) Cleverer people know this. Not so clever people are putting too much faith in their formulas because they worked before, when things were quite different.

kao_hsien_chih

Based on the numbers that I'd worked with, I'd give Trump about 1/3 chance (not very educated guess. I don't completely trust the numbers I worked out myself--too much uncertainty.) of carrying the electoral college majority, although much less that he'd carry popular plurality (but not the kind of probabilities that poll aggregators are chirping about). The votes just aren't there to make a Trump majority too likely, but there are too many strange things going on to stake too much on one conclusion or the other. The assumptions and formulas that pollsters seem to be using are just too inappropriate for the present state of things. If I were a betting person, this would be excellent betting odds: Trump is an underdog by a decent margin, but not nearly the underdog that the polls seem to suggest he is.

Fred

Unleash the Sardaukar!. paid off polling companies. I believe the bias is more manipulation of the electorate. It's pretty obvious at the bottom of the poll you link too (Zeitgeist heading) that it is being (perhaps) effective in discouraging actual Trump voters so they stay home on election day (as if there were no other candidates or issues on the ballot to vote for or against). Clinton is obviously worried in the drop off in support (compared to Obama's election) of the 18-44yo portion of the electorate. The very high level of support for third party candidates is very abnormal since that represents about a significant increase over the max seen by anything other than a nationally known candidate (Perot, Nader).

Paul Escobar

Mr. Lang,

It is October 23 2016.

So in regards to IBD/TIPP...
Let's look back at their performance, at the same time period in late October 2012.

Two interesting articles. One from a liberal source, the other from a conservative source.

---

From the the liberal Huffington Post, dated October 25, 2012:
http://www.huffingtonpost.com/2012/10/25/presidential-polls-romney-surge_n_2016066.html

The liberal author addresses the hoopla about a "Romney Surge". He describes Obama leading in state polls, but trailing Romney in nationwide polling.

On October 25 2012, the following nationwide tracking polls put Romney in the lead:
- ABC/POST
- IPSOS/REUTERS
- GALLUP
- RASMUSSEN
- AP/GfK

On October 25 2012, who has Obama ahead by 3 points?
- IBD/TIPP

As you know, this would end up being very close to the actual result. And I believe IBD/TIPP's *final* result came even closer in accuracy.

---

From the conservative PJ Media, dated October 24th 2012:
https://pjmedia.com/blog/polls-with-large-voter-samples-all-favor-romney/

To explain away IBD/TIPP's result, this conservative author assures us there's a problem with their sample sizes.

He contends that large samples tend to show Romney ahead. While smaller boutique samples (like IBD/TIPP's) show Obama ahead.

He reminds his readers:
"Statisticians will tell you that the larger the sample size, the more reliable the poll. A lower margin of error, and thus a greater level of reliability, for the large-sample pro-Romney polls. Each and every pro-Obama poll has a higher margin of error, and is thus less reliable."

In his conclusion, IBD/TIPP showing Obama +3 over Romney was "less reliable".

---

It is also interesting to note, Mayur of TIPP & the IBD staff are staunch conservatives. So all the more impressive that they would be an outlier in regards to Obama, back at that moment of time in 2012.

Best,
Paul

The comments to this entry are closed.

My Photo

August 2020

Sun Mon Tue Wed Thu Fri Sat
            1
2 3 4 5 6 7 8
9 10 11 12 13 14 15
16 17 18 19 20 21 22
23 24 25 26 27 28 29
30 31          
Blog powered by Typepad