October 11, 2015

Poll Watch: Trump Rebounding, Rising Back Toward the 30% Level

  10:10 pm

After attacking Carly Fiorina and getting gutted during the Simi Valley debate, Donald Trump’s numbers came crashing down. Having peaked in the low 30s, The Donald saw his numbers nationally and in many states dip into the low 20s – and even fell as far as the high teens in a handful of polls.

Last week, it was becoming obvious that Trump had managed to stop the bleeding and was leveling out. This week, it is becoming apparent that Trump is once again on the rise, and this CBS News poll is the latest piece of evidence pointing in that direction:

CBS News National Republican Primary

  • Trump – 27%
  • Carson – 21%
  • Cruz – 9%
  • Rubio – 8%
  • Bush – 6%
  • Fiorina – 6%
  • Paul – 4%
  • Christie – 3%
  • Huckabee – 2%
  • Kasich – 2%
  • Santorum – 1%
  • Graham – 0%
  • Jindal – 0%
  • Pataki – 0%

Survey of 419 Republican primary voters was conducted Oct 4-8 and has a margin of error of ±5%.


Sunday Open Thread

  5:00 am

This is the Open Thread for Sunday.

City of the Big Shoulders: They tell me you are wicked and I believe them, for I have seen your painted women under the gas lamps luring the farm boys.

This is also a good place for new polls or articles you think might be of interest.


October 10, 2015

More on Internet Polls and the Trump Effect

  6:31 pm

A few follow-up thoughts on Matt Coulter’s post yesterday

– Yesterday’s newsletter from HuffPost Pollster was dedicated primarily to discussing the fallout from Gallup’s withdrawal from the horserace polling biz (btw, Pew has also announced they are cutting back). A variety of opinions worth study.

— A major point Matt made was about the huge variance between live polls and internet polls in regard to Donald Trumps performance. In the comments, a Trump supporter, Gerry, wrote thus:

Trump does better with automated phone and internet because people don’t want to tell a human being they are voting for Trump

Whether it’s true or not in this case, there’s a long history of related phenomena: the Shy Tory Factor, the Bradley Effect, the Spiral of Silence. More simply, I remember an advertising exec who spoke to my high school class telling us that if we believed what people said, most people were watching National Geographic documentaries on PBS rather than Gilligan’s Island. (Yes, I’m old).

If there is a Trump Effect(tm) this time, then that might mean that the online polls are more accurate than I an inclined to think this year.

– In related news, the Los Angeles Times works with one of the best ‘real’ pollsters in the business – Field. They have consistently resisted online polling. Guess what? They are now experimenting with it.

The Los Angeles Times is releasing its first ever online poll, a snapshot of residents in our community and their thoughts about their neighborhoods, and the advantages and disadvantages of living here.

The USC Dornsife/California Community Foundation/Los Angeles Times poll was conducted Sept. 10-24 of 1,500 adults who live in Los Angeles County.

So what’s the big deal?

Dating back decades, we’ve done phone polling, in which surveyors randomly dial phone numbers from a sampling of registered voters and pose a series of questions to people who answer and are willing to participate. … A successful interview can last anywhere from 15 to 30 minutes. Fewer people have land lines today, and resistance to participating in telephone polls has increased. Phone polls have long been considered the standard, but the polling industry is in flux. Some of the top organizations doing the surveys say costly and time-consuming phone polls are no longer the best method.

I don’t have any answers, I’m not sure I even know the right questions yet; and I don’t expect to find the answer in today’s discussion, or anytime soon. This is something that we are going to be discussing (here and elsewhere) for, probably, years.


Saturday Open Thread

  5:00 am

This is the Open Thread for Saturday.

Mama pajama rolled outta bed, she ran to the police station; when the cop found out, he began to shout, he started the investigation

This is also a good place for new polls or articles you think might be of interest.


October 9, 2015

Poll Watch: Four Republicans Beat Hillary in NC, Trump Loses

  6:25 pm

North Carolina is, of course, an absolute must win for the GOP in November — and it looks like it will be solidly red unless Trump or Cruz end up being the nominee:

Civitas North Carolina General Election Matchups

  • Carson – 50%
  • Clinton – 40%
  • Rubio – 47%
  • Clinton – 42%
  • Bush – 45%
  • Clinton – 41%
  • Fiorina – 45%
  • Clinton – 42%
  • Clinton – 44%
  • Cruz – 42%
  • Clinton – 47%
  • Trump – 43%

Survey of 600 registered voters was conducted Sept 28-30 and has a margin of error of ±4%.

Also of interest was this question included in the survey:

trump civitas nc


Poll Watch: Trump Back to Double Digit Lead in New Hampshire

  3:43 pm

Trump is at his highest level of support in New Hampshire since the beginning of September, at least according to Gravis. Hat tip to GS for the numbers:

Gravis Marketing New Hampshire Republican Primary

Trump 32% (32%)
Carson 13% (8%)
Kasich 10% (15%)
Rubio 8% (3%)
Fiorina 8% (2%)
Bush 8% (7%)
Cruz 5% (3%)
Christie 3% (9%)
Paul 2% (6%)
Graham 1% (2%)
Santorum 1% (2%)
Jindal 0% (2%)
Huckabee 0% (1%)
Pataki 0% (1%)

Automated phone survey of 662 Republicans was conducted Oct 5-6 and has a margin of error of ±3.8%.


On Gallup, Internet Polls, and the Future of the Polling Industry

  1:03 pm

Gallup dropped a bombshell on the political world earlier this week when they announced they would not be conducting any horserace polls for the primaries during this campaign, and probably wouldn’t be doing any for the general election, either.

One of the most vaunted and trusted pollsters for decades, Gallup’s decision became by far the largest indication yet that something is amiss in the polling universe. Pundits have called the decision “stunning” and “surprising” – and Harry Enten over at 538 didn’t mince words when he declared “that sucks.” After a respected hisotry of accuracy, Gallup missed the result of the 2012 general election by 5 percent (they had Romney +1 instead of Obama +4). Following that, Gallup underwent a detailed internal and external investigation to determine what went wrong. The result of that investigation, in not so many words, was this: it’s too expensive and too time consuming to properly conduct polls in this day and age.

That conclusion is the same one already reached by other pollsters and pundits, but having Gallup’s weight added to the pile is a clear warning sign to the rest of the industry. Here’s the gist of the problem: Gallup’s investigation determined that live telephone interviews were, hands down, the most accurate method of polling. Unfortunately, live telephone interviews are also the most costly and time intensive method of polling. In a world where fewer and fewer people have land lines, pollsters should be calling more and more cell phones — something that is inherently more difficult and more expensive because of federal regulations. Additionally, studies – including Gallup’s – are showing that fewer and fewer Americans are willing to even answer pollster’s calls and/or take part in a survey; therefore, many more calls than ever before have to be attempted in order to get a similar sample size. Now in 2015, it’s becoming a case of diminishing returns. Whereas sample sizes in the 2012 and 2008 elections were commonly well over 1,000 participants and rarely dipped below four or five hundred, sample sizes of 2016 surveys are the opposite: rarely more than three or four hundred. Many of the national surveys we’ve seen sport sample sizes around 200 or 300 – a standard which would have been laughable at four years ago. Today, though, it illustrates the increasingly inherent difficulty of reaching respondents.

And so, pollsters have, it seems, four options moving forward here in 2015:

  1. Bow out from horse race polling all together, as Gallup has. Others, like Pew Research, are already following Gallup’s lead in this way, and experts predict others will follow suit as well.
  2. Conduct traditional, accurate polling using live telephone methods but pay more for statistically significant sample sizes.
  3. Conduct traditional, accurate polling using live telephone methods but accept much smaller sample sizes, leading to larger margins of error and less statistically significant results. This is what most pollsters appear to be choosing at the moment, because, let’s be honest: the general poll consumer isn’t educated enough to care about things like sample size and margin of error.
  4. Conduct internet polls.

Many have ventured out into realm number four, online polling, and they are driving some headlines. Organizations like Morning Consult, Reuters, Survey Monkey, and YouGov are pumping out online polls much more consistently than traditional phone-based pollsters can – but the burning question remains: how accurate are they? And an even bigger question, at least in the minds of poll-watching nerds and statisticians, is: can they even be considered statistically scientific surveys?

The answer, thus far, seems to be a resounding “no” on both accounts. Internet polls seem to provide results wildly different from traditional phone polling. Consider, for instance, the Pollster.com average of only phone surveys for the 2016 Republican primary race:

phone polls only

Now compare that to the same average but using only online surveys:

internet polls only

It’s like they’re polling two entirely different races. Is Donald Trump averaging 18% nationally and leading Ben Carson by just two points (phone surveys), or is he averaging over 30% and leading everyone by 16 points (online surveys)? Obviously, both cannot be true, and when given the option of which one to believe, Gallup, along with statisticians everywhere, would point us toward traditional phone surveys. Indeed, over at 538 they note, “Polling consumers are far better off in a world of Gallups than in a world of Zogby Internet polls.”

However, that is certainly not to diminish the efforts of the enterprising organizations who are embarking on internet polling quests this year. The polling world is necessarily undergoing some changes at a fundamental level, and pollsters are going to have to figure out how best to move forward. The good news is that this has happened before. Polling used to be conducted, believe it or not, via mail. When personal landline phones became more mainstream, there was a period of awkward transition as pollsters attempted to shift away from mail surveys and toward phone surveys. This led to several infamously inaccurate polling results (Dewey defeats Truman, anyone?) as pollsters tried to figure out how to get a truly representative, random sample based on telephone usage. (In the beginning, for instance, rich people who tended to be Republicans were much more likely to have a personal phone, skewing poll results.)

Eventually, of course, pollsters perfected the art of telephone surveys and they became the gold standard. Now, as landline phone usage and people’s willingness to talk to phone pollsters both decline, pollsters are having to switch over to the new medium of polling: the internet. While they make this switch, there will, again, be a period of awkward transition as pollsters attempt to figure out how to get a truly representative, random sample based on internet usage. That is what we’re presently living through, and at the moment, that standard is impossible.

Right now, all internet surveys are necessarily opt-in, self-selected surveys that cannot be random or representative – two necessities to be at all meaningful for statistical purposes. How do pollsters adjust for this fact? By utilizing extensive and heavy weighting to manipulate their samples. Weighting is already a delicate subject when it comes to telephone surveys (kind of a necessary evil, if you will, which can lead to poll herding — a fascinating topic for another post); when it comes to internet surveys, weighting presents a sort of uncharted frontier where pollsters are exploring unmapped territory. Twenty years from now, the general consensus is that all polling will be done via the internet. We will have made the switch, just as we made the switch from mail to phone decades ago. Until that time, however, pollsters face some impossibly difficult decisions regarding how to conduct surveys in the present – meaning that the polls we’ve come to know and love most likely aren’t nearly as accurate as we would like to believe they are.


Friday Open Thread

  5:00 am

This is the Open Thread for Friday.

He was an old man who fished alone in a skiff in the Gulf Stream and he had gone eighty-four days now without taking a fish.

This is also a good place for new polls or articles you think might be of interest.


October 8, 2015

A House Divided Against Itself Cannot Stand

  2:53 pm

Kevin McCarthy of California, House Majority Leader, dropping out of the race for Speaker of the House on the day of the election among the Republican Caucus merits a third FPP on the contest this week. The first two were predicated on the basic math that more than suggested that McCarthy couldn’t get to 218 votes, something absolutely necessary for a Republican to prevail in the election involving the entire chamber .

This basic math was due to intensely strong feelings on the part of those willing to go to the mattresses against the old leadership. While there are 247 Republicans in the House and only 188 Democrats, the fact that there are more than 30 members with those feelings necessitates finding someone outside of the old leadership group to represent the Party.

This problem seemed unsolvable until Jason Chaffetz of Utah threw his hat into the ring, running on the idea that someone with the right disposition, and with superb communication skills, and an established record of working well with both sides is needed to bridge the divide and establish unity.

Having watched the clueless reaction by both Fox and CNN following McCarthy’s decision, it’s safe to say that his election had been assumed to be inevitable, and the brief news flashes of Chaffetz getting into the race had been mere lip service to his announcement.  But it is clear that Jason’s reason for getting in, that the math didn’t add up, was right on target.

Meanwhile, the obvious people to take on the job have all declined to go after it.  Paul Ryan of Wisconsin wants to stay on as head of Ways and Means.  Jeb Hensarling of Texas and Trey Gowdy of South Carolina are going to stay on as head of their respective committees.  No major figure has stepped up or expressed interest as of now.  Daniel Webster of Florida, who was in the race, and who had won the support of the House Freedom Caucus led by Jim Jordan of Ohio, has expressed that he’s not sure if he will stay in  with the majority leader out of it.  He was, by implication, not really running for the job as much as he was standing in opposition to the status quo.

Only Chaffetz, the head of The House Oversight and Government Reform Committee, is definitely still in the running,  and he is still campaigning and lining up support.  Unless someone with real stature in the body jumps in, it’s hard to see anyone else coming out of this.  Chaffetz is the logical candidate to gain support from both sides.  If he does, the House will get more combative, and more conservative.

He figured out the math before anyone else and offered a solution to it.



Field Poll: California Primary

  2:46 pm

From Field Poll’s write-up (with full results and methodology):

Three political outsiders with no prior experience in political office are now leading in California among Republicans likely to vote in the state’s June 2016 presidential primary. Businessman Donald Trump, at 17%, and physician Ben Carson, at 15%, and businesswoman Carly Fiorina, at 13%, are besting the crowded field of current and former U.S. Senators and Governors, who collectively receive 42% of the vote. Another 13% are undecided. […]

Despite the fact that Trump has been leading in most nationwide polls for more than two months, relatively few likely Republican voters in this state (28%) think he will eventually become their party’s standard-bearer. Also, when California Republicans are asked what their personal reaction would be should the billionaire businessman win the GOP nomination, greater than four in ten (44%) say they would be either upset or dissatisfied, and 37% are not confident that Trump would be able to defeat the eventual Democratic Party’s nominee.

Calif poll
Sample 214 likely Republican primary voters. Dates: 9/17-10/4. Margin of error: +/-7.

As an aside, if you take a few moments to read the Field Poll’s methodology, it is pretty much the ideal way for polls to be conducted. Unfortunately, it is also expensive, so I doubt that it will be widely adopted (though a few others, such as Seltzer, use similar methods).

H/T: GS.


Recent Posts

Search R4'16