Every series of polls has the occasional outlier and in our opinion this is one....Here's how Real Clear Politics displays the recent polling. I've framed the screen grab to highlight recent Reuters polls (click to enlarge):
The Reuters/Ipsos Core Political poll has a significant realignment this week across a number of metrics. Most pronounced is President Trump’s approval rating which currently sits at 48% with all Americans. His number with registered voters is essentially the same at 49%. Corresponding with Trump’s stronger approval rating, evaluations of his job performance across the board are stronger this week from 57% approving of his handling of the economy to 44% approving of the way he treats people like them.
There was a -19 spread between approval and disapproval in the second week of April. The spread a week ago was -14. It is now 0. The official Reuters position on that is: Sometimes you get an outlier. We'll show you the numbers, "in the interest of transparency," but our theory is, that's an outlier.
Posts
79 comments:
They both look like outliers. Without all the polling parameters and protocols, it is impossible to know if is just an outlier or whether there is a flaw in the way the poll is taken.
Outlier? Perhaps. But looking at the RCP polling average history, there is a trend and that trend is definitely up for Trump.
If it is an outlier, that is, not true, then the poll itself is flawed. What are the flaws that lead to it being not true?
-sw
I think that's fair. Which in this environment is an achievement.
Of course it's an outlier. It's favorable to Republicans. Someone needs to fix it.
What are the flaws that lead to it being not true?
Not necessarily flaws. Maybe just randomness--in people's opinions or in the mix of people who happened to be sampled. A cold day in July is an outlier but it happens (I remember a very disappointing one).
Maybe an outlier. Maybe they got it right for once.
What changed?
Kanye, baby.
We want more Stormy! Nothing spells success like endless coverage about a one-nighter with a minor league porn star 12 years ago.
That old Giuliani magic--America's Mayor wins hearts and minds.
The press really is the enemy. If they say it's black, it really is white, and vice versa.
Trump mind-control. He will bend you to his will. You are powerless to resist.
Comey does it again. The one-man wrecking crew lumbers around the national stage, boosting Trump through constant displays of Comey-sized ego and preening. He truly is the only honest man in Washington--give me more Trump!
Didn't the same polling companies say that Hillary had a 98% chance of winning?
More Stormy! That will bring him down! She said she was a supporter, maybe she is playing eleven dimensional chess too!
I used to roll my eyes when my wife was watching some entertainment program purportedly about the entertainment business and celebrities and Stormy kept coming on, now I cheer them on.
The reason it’s not a flat out outlier is because it is part of a trend. And you know what? Maybe people are reacting to the result of Trump’s diplomacy on the Korean peninsula, and comparing it to the predictions of doom and almost Hillary level ham-fisted warmongering the media had, where they got it one hundred percent wrong, just like the economy, the stock market, unemployment, etc, etc, etc.
How many times can somebody with the bullhorn the media has be precisely and emphatically wrong without people noticing?
Bear in mind we vote by states. Trump will lose CA and NY by large margins but I can see him winning re-election.
Outliers are suspected when there are no apparent causes. Trump has possibly ended the fucking Korean War.
So looking at world events that could have moved the polls since their last poll (-14).
- North/South Korea declare peace. Trump credited (calls for Nobel)
- Kanye West declares support for Trump, ripping hole in space/time continuum.
- Good Economic data.
Hmmmmm
Coming up next week--a Reuters Ipsos poll of 10,398 RV--they sample 8400 additional people from California, come up up with a Trump -61 poll, and decide the new poll is not an outlier--it reflects the way people should feel.
f it is an outlier, that is, not true, then the poll itself is flawed. What are the flaws that lead to it being not true?
A 95% confidence means that one in twenty are wrong. But for the reasons outlined in this thread by various parties, it may have had a more Trump friendly sample than average, but one need not go looking very far for reasons to support a higher approval, unless, like for example, the Washington Post, your confirmation bias is so pernicious that you are unable to perceive reality.
So, is it possible that form a strong plurality that
Tax Cuts + Protectionism + NK Appearing to soften + Iran Deal Lies + FBI Corruption + Media/Hollywood/Dem TDS waving loud and proud >> Russia + Mueller + Manafort + Stormy Daniels (I am not even sure which side to put Parkland or DACA)
Yes, yes it is.
And in random samples, some will be weird, In fact, the fewer weird results, the more the data is manipulated / biased.
And Maybe the calls from Dem leaders to stop the impeachment rhetoric is an outlier as well.
One cannot rule out flaws of course. These polling firms have agendas in public polls, basically a public polling firm is like the proverbial mushroom farmer, they feed the public bullshit and keep us in the dark. The press mostly uses the polls as instrumentation for their narrative. Is it taking? Do we need to adjust our tactics? How is Stormy playing? Should we switch to some other story?
Personally, I agree with CNN’s news judgement that the forcible rape of Juanita Broaddrick justified zero coverage, and that Trump’s consensual tryst with a gold digging porn star is more important that the potential end of a war that has been going on since the fifties and occupied the attention of US administrations for longer than this 60 year old has been alive.
Think about what the numbers would be if Trump didn't have the media, Mueller, the Dems, late night comedy, daytime TV (the View), AA comedy (Inga, LLR, Ritmo) nipping at his heels 24 hours/ day.
Think about what the numbers would be if Trump didn’t have the media, Mueller, the Dems, late night comedy, daytime TV (the View), AA comedy (Inga, LLR, Ritmo) nipping at his heels 24 hours/ day.
Oh, they have, and they shit their knickers.
I guess it shows if you tell enough lies and distract the focus from the lack of ethics in the swamp people will be fooled some of the time.
Did I leave out the swamp dwelling R's? Why yes I did.
You can't fool all the people all of the time, but if you can do it just once, it lasts for four years. - Roger Price
You can use the word outlier to refer to something unusual and counter to observed and documented events, such as the cold day in July example above. You can also use outlier to categorize something as false, worthy of discarding, which is what is done in this reporting. Thus we are informed of the reporter's bias.
-sw
The 2016 election was far from a fluke
"I guess it shows if you tell enough lies and distract the focus from the lack of ethics in the swamp people will be fooled some of the time."
It's worked for the Democrats in Illinois for years. Why change a winning formula?
I started responding to telephone polls last week. Maybe others who are fed up with all of the negative reporting did too. I hadn’t responded to a poll in years.
The trouble with polls is that they don't reflect serious opinion. It's any old answer given without thought or commitment.
A vote, unlike an opinion, involves a moment of madness where you say yes to your choice in order to make it yours.
Maybe this Poll asked more Koreans for their opinion. They are strongly Christian and want a strong leader trusted enough by all to finally end the 68 year old war Harry Truman started when he refused to surrender to Stalin and Mao.
@ brylun @ 7:43
Thanks for that link! A worthwhile read and I will be buying the book.
1960's tech analogy
If you set up to measure the temperature in the basement, you get readings like 65.3 64 62.7 67.3 99999903.4 69.0 64.2 63.2
averaging isn't a good idea. Median is a good alternative.
Kanye
Comey
Stormy
North Korea
Tax time?
There are several issues at play.
First, there is the sample frame that is used to generate the polling call list. This is a complicated business and very easily sabotaged in the direction the pollster wishes to go. At primary risk, so to speak, are people on the original sample who are not at home. Does the pollster have a separate random list of names to replace THAT person or do they go to the next person on the list. Many cheap pollsters take the second option the first being too expensive.
Second, in such polls the actual question is not the TYPE I error (as in the confidence interval of 95% which is the measure of TYPE I error classic). The issue to be asked properly is what if the poll is wrong and the results are not what are reported. We see statistical significance but maybe we are wrong. This is TYPE II error that essentially asks, "What if we are wrong?"
TYPE II error has to addressed when the Sample Frame is being constructed and requires sensitivity analysis sort of like a Bayesian Process. Complicated and too much time to do. Most Pollsters are gambling that there is neither such an error AND the readership is too stupid to know the difference.
Finally, polls in general are like Sports Announcers who laud the team with the ball: whoever has the ball is the victor. Bullshit.
The fact is that the current political APPEARS to be in play and each successive poll will convince the (stupid deplorable) electorate to change their mind or that they just don't know the situation.
Methods to document CHANGE in trends are not analyzed by a simple average. That is some bullshit English Majors with no stats or method design think is appropriate. Since the larger political situation is very much in change and anything but static, the traditional dumbshit pollsters have to do more work, cutting into their profit margins, to get better results.
The real pity in the polling industry is that they generally lack the balls to tell their clients that the world does not and will not conform to their hypotheses about the electorate and what they REALLY TRULY want.
This says nothing about their inability to look at the numbers objectively rather than through their Hillary bifocals.
Gangster in an old Bugs Bunny piece: "I don't know how youse done it, but I know youse done it!" Presidential approval polls and generic Congressional ballot both looking better for Trump as the year goes on.
Trump's successes at least give off hints of a re-alignment of parties, based on the failure of either party to deliver to voters. His skills may lie more with putting a spotlight on the old, rather than building something new, but he is definitely accomplishing something. As for Korea, yes, if Obama deserved the Nobel Peace Prize for doing more or less nothing, and Al Gore ahem, cough, never mind, then it may be reasonable to consider Trump as a possible co-winner.
Whenever I'm polled about anything, I just make up shit. I figure I'm not under oath and have no obligation play their game.
Maybe our silent army of minor trouble-makers are skewing the data.
guess it shows if you tell enough lies and distract the focus from the lack of ethics in the swamp people will be fooled some of the time.
So the stock market isn’t up huge since the election, unemployment at record lows, historic lows for minorities, blacks, included. It isn’t looking like the Korean War may be ending after dragging on since before almost all of us Althouse alte kakers wer born?
Is that all “lies”? Or are you suffering from pernicious confirmation bias?
For Reuters to claim an outlier is very misleading in terms of the probability of their poll. Let's say that Reuters believes that the difference in the approval/disapproval number is closer to -14% than zero, which were the last two results.
Assuming that the polls are truly a random sample in both cases, the margin of error is about 2.6% for a 95% confidence for the sample size of the polls.
For the most recent poll, that means Trump could have a net approval of between +2.6 and -2.6% at a 95% confidence. but Reuters obviously believes that the net approval is somewhere closer to -10%. If this poll is random, to get the spread to include +5.2% to -5.2% is a confidence of 99.995%, or 1/20,000, meaning only one poll out of 20,000 would get this result if the actual spread was -5.2%. To get to +/- 10%, it is 1/20 trillon. So either their poll is entirely untrustworthy, or Trump has a better approval than they think.
That is some bullshit English Majors with no stats or method design think is appropriate.
Stop bashing us English majors! OK, I have had some coursework in stats probabilities, and discrete math. I would think that exponential smoothing would be the way to handle trends in polling, which would limit the impact of true outliers. Statistical process control. It’s a well known thing.
What might create an “outlier” could be that they failed to ask the question in such a way as to produce the desired answer too.
Mueller went after Trump's attorney in NY.
Mueller's questions for Trump came out.
The WHCD fiasco happened.
Trump punished Assad for Chemical weapons.
Trump seemed to get on very well with the new French president.
Roseanne Barr
brylun at 7:43 AM
The 2016 election was far from a fluke
Thanks for the link to that article.
If the Leftist Collectivists try just a bit harder to think of PURPLE ELEPHANTS the American people will try diligently to give a damn about the things Democrats have spent months telling them are important.
The PURPLE ELEPHANTS cannot be unimagined.
As a left-leaning organization, Ruetera skewed the poll and still got 48% for Trump.
PURPLE ELEPHANTS, punks.
Shorter Reuters: Trump’s numbers are starting to look good in this, an election year. THAT can’t be right.
Look at the Gallup poll, which is running about half a week behind the Reuters poll.
Gallup shows an 8 point rise from -19 to -11
What is suddenly changing may not be people's opinion as much as their willingness to express their opinion.
Remember that WaPo reporter whom they had to fire because she was outed giving a talk at an activist conference, one aspect of which was the use of public polling to gauge the effectiveness of their tactics and to refine their narrative?
What is suddenly changing may not be people’s opinion as much as their willingness to express their opinion.
Roseanne and Kanye? What a country!
There is very likely a 5-15% under reporting in this poll. Conservatives don't like or trust Academic pollsters or the media and like as not hang up on these odious windbags. Plus Republicans tend to be at their JOBS.
Trump has no path to 270 electoral votes in 2020!
Kanye for a long time only followed Kim Kardashian on twitter. He recently started following Candance Owens and Emma Gonzalez. It will be interesting to see where this leads.
Candace, not Candance. Not sure if she can dance or not!
The other thing about the constant “Trump is a liar” drumbeat is that Trump’s predictions on the big things, and even the Russia persecution have borne out over time, not the hallmark of a liar.
Trump has no path to 270 electoral votes in 2020!
I miss America’s Politico!
@brylun, thanks for the pointer. I’ll order the book when I order Chris Buckley’s latest — using the Althouse Amazon portal as always.
Liar Hillary was defeated. The liar left must march the butt hurt parade down the street forever.
Verrry interesting. Obama got 75%+ positive media coverage. Trump gets 75%+ negative coverage. Approval ratings after one year - pretty much the same.
"What does it tell you, Tonto?"
"It tells me somebody stole our tent, Kemosabe."
What it tells me is that the media don't matter. They're IRRELEVANT.
(In Navajo, on the other hand, “kemosabe” translates as “soggy shrub.” If this seems an odd thing for faithful friend Tonto to call the Lone Ranger, perhaps he was just repaying the Ranger's long-standing insult. “Tonto,” after all, is a Spanish word meaning “stupid.”)
Verrry interesting. Obama got 75%+ positive media coverage. Trump gets 75%+ negative coverage. Approval ratings after one year - pretty much the same.
What it tells me is that the media don't matter. They're IRRELEVANT.
Or it could mean that if reversed, Obama would have been at 40% approval and Trump at 60%.
I remember Instapundit linking to a study a few years ago that said if the media was truly neutral, the average US voting pattern would resemble Tennessee.
Shorter Reuters: We can't believe the anti-Trumpers aren't up by 50 points!
That keeps them from having to report on it this week. It keeps them from having to explain the tax cuts are working, employment is up, and the Mueller investigation has revealed itself as a witch hunt.
The last thing they want to do is a series of reports on why people are starting the favor the president.
This buys them a week to adjust the polling to something like D+75.
Roy Lofquist,
The PURPLE ELEPHANT strategy has one purpose: make the press irrelevant.
The press cannot stop talking about PURPLE ELEPHANTS. And then they ask why Trump keeps tweeting about PURPLE ELEPHANTS.
The press cannot see what Trump sees: the press is the enemy. Their ability to control the narrative (read: the lie of the day) moves public opinion at the margins. So degrading the press' influence is paramount.
Whoa. That's quite a shift. Even an outlier would be unlikely to be that far out.
And if the poll trend was reversed, would they still be calling it an outlier, or saying "Trump popularity plunges!"
By calling it an outlier, they are performing a public service. Reporting it straight would result in mass hysteria. Best to let the TDS crowd stay in their bubble, where they believe Trump won't even last till 2019.
Power Line has the stats from the poll. Essentially it was a balanced poll - equal #s of Rs and Ds. The first time that had happened in weeks; so, naturally, it is an outlier which they will correct next week. Kevin had it right!
Look at the trend lines over time and the margin of error. A poll is not a psychic prediction or a guarantee.
I suspect the commenter who wrote it might indicate that some Trump supporters are finally willing to start answering a public poll. Polling approval/disapproval for a figure like Trump isn't like polling the question, "What is your favorite color?" There is no sense on the part of the respondent that the person asking the question will dislike one of your answers a lot more than the other.
I don't answer telephone polling questions any longer, but I did at one time in my life. During the Monica Lewinsky scandal in 1998, I received a phone call from a polling agency polling the question (paraphrased), "Do you believe the President's denials about having had sex with Lewinsky?" At the time, the affair was only a rumor and Clinton had denied it publicly, but this was before the existence of the blue dress had been reported. I responded that I did not believe Clinton's denial. The person making my call then proceeded to rephrase the question three different ways, the gist of which was trying to get me to say I believed Clinton's denial- in essence, she was arguing with me. I hung up after the 4th iteration.
I have long suspected that Republicans/conservatives reply at a lower rate than Democrats/liberals to political polling. The good polling companies probably do try to correct for this problem to some degree, but it wouldn't take much of an error to produce outlying and revealed results. Just having a non-response differential of 5 people in a 100 between Trump supporters and Trump detractors would be enough to produce a -6% spread in a poll where the true result is 50/50.
How hard to you think some of these polling organizations try to correct for obvious non-repsonse problems? Not very hard would be my guess in a lot of these cases. One of the interesting things about this "outlier" is this- the poll had a balance of self-identified Rs vs Ds- something that was different than the polls showed a few weeks back. Is it because the poll suddenly was over-sampling Rs? Or is it because Rs are suddenly less prone to hanging up or not answering? Who knows?
How hard to you think some of these polling organizations try to correct for obvious non-repsonse problems?
Many years ago I took a course on Survey design. It was part of a Masters program in medical quality research and improvement.
Low income people are very unreliable on surveys because they will try to answer with the answer they think the surveyor wants.
I don't pay a lot of attention to polling and never answer such calls.
It sounds like you encountered a "push poll."
If the margin of error is +/- 3% the significant digit is in the tens place.
This looks like them begging people not to shoot the messenger.
Yet, all three of these points, I've seen the media desperately try to discredit.
"North/South Korea declare peace. Trump credited (calls for Nobel)." -- That this work is all being done by North and South Korea, and Trump is uninvolved is a major meme.
"Kanye West declares support for Trump, ripping hole in space/time continuum." -- The de-personing of Kanye West is happening at a rapid pace. Was Chris Brown, who abused someone, as aggressively de-personed as Kanye?
"Good Economic data." -- Much like homelessness, now that a Republican is in office, the media remembers there are other indicators besides raw unemployment.
Yes, Prime Minister is almost always relevant.
In the context of the RCP average, no. It is not an outlier. 3 of the last 5 polls have him within the margin of error. The other 2 have him down significantly, but much better than earlier polls. So solidly on one side of the bell curve, perhaps, but not on either tail.
@ Matthew Sablan Exactly!
I think Roseanne and Kanye broke the inhibitions of the shy Trump voter.
Polls:
"It has become increasingly difficult to contact potential respondents and to persuade them to participate. The percentage of households in a sample that are successfully interviewed – the response rate – has fallen dramatically. At Pew Research, the response rate of a typical telephone survey was 36% in 1997 and is just 9% today."
http://www.people-press.org/2012/05/15/assessing-the-representativeness-of-public-opinion-surveys/
The polls are less than useless - they have been weaponized.
“Verrry interesting. Obama got 75%+ positive media coverage. Trump gets 75%+ negative coverage. Approval ratings after one year - pretty much the same.
"What does it tell you, Tonto?"
"It tells me somebody stole our tent, Kemosabe."”
The Free Stuff floor of approval can be raised but it generally cannot be lowered, except in unusually good economic times. Which, intentionally, the Obama years weren’t.
Yes, I agree, it does definitely look like an outlier, a bad one. But, the good news is, the more recent numbers show that this condition has improved.
There's outliers and out-and-out liars. Pick your poison.
"There was a -19 spread between approval and disapproval in the second week of April. The spread a week ago was -14. It is now 0. The official Reuters position on that is: Sometimes you get an outlier. We'll show you the numbers, "in the interest of transparency," but our theory is, that's an outlier."
When plotting a graph, you need more then one data point to see movement.
One dot shows you a lot less then 2 dots.
Two dots shows you less then 3 dots.
Get a couple of dots, and you can see trend lines.
It could be an outlier, but the Rasmussen poll also shows a positive +2. And if you click on the link the Rasmussen polls, they've generally tracked with the other polls at RCP? (Like March 30th was at -5 for Trump).
The Strongly Disapprove number as Rasmussen haven't changed in months, staying between 39% and 42%. The Strongly Approve number have come up a little. So it looks like the change is in the weakly approve and disapprove middle ground.
If you look at the Reuters polling results presentations for this poll and the last, it looks like this poll grab more Moderate Republicans than usual. But it also grabbed more Moderate Democrats (though not as many).
One thing under-noted here is that this poll and all the others are NOT Likely Voter screens yet. Reuters shows aregistered Voters which is a tad closer to Election Day reality than All. If a Republican is even in Register avoter polls, he or she is up 5 on Election Day.
Dem waves, dem waves, dem Blue waves
Bout to hear the voice of the Lord.
How many outliers does it take to make a trend?
Post a Comment