Then there are the millions and millions like me who haven't voted in twenty years or more who are voting this time just because Trump is what we've been waiting for. The unlikely,unpolled voters.
9
Thought you were going to say that the U.S.C./LAT poll overlooked people with only cell phones. I'd be Hillary is polling 2:1 in that demographic. In fact, Trump polls behind the Libertarians and the Greens in the youngest demographic.
15
Comparable poll in 2012 by RAND was one of two polls to accurately predict Obama's margin of about 3.5%. All others failed. The other poll was IBD/TIPP - currently showing a Clinton lead of 7.
11
I am no expert on polling, but I do know human nature. The USC poll seems to me to exhibit the desire to do some clever up-front work, put the system on auto-pilot and reap the results without much additional effort. As in most things the quality of the product is usually a measure of the effort.
12
In some respects it's difficult to believe that Trump was leading two weeks ago and that it is still this close, even though he might be losing by as much as 15 pts everywhere. The only way these polls can be permanently right is if the crazy week has been an epiphany point on Trump for 15-20 of actual voters. They would be saying something like I might want change of the Trump kind but unfortunately he is the wrong guy to deliver it, so reluctantly I'll stick with more of the same for now.
4
Last British election (not Brexit referendum - last poll was 52/48 for Remain, result was the opposite) polls were massively wrong due to shy Tory factor. The polls predicted a Labor win. Result was a clear Tory majority, well beyond Tories wildest expectations. People lied to pollsters about real intent to vote Tory. USC/LA Times poll is based on most successful 2012 RAND poll that accurately predicted Obama win by 4% (actual 3.9%), when others were at about 0.7% or worse.
3
we are currently experiencing the Bradley Effect here in the US. bradley effect is named after a black politician who in every prediction in all the polls seemed to have a clear easy win, but ended up suffering a massive defeat on election day.
it was studied that people were more likely to lie and say they were voting for Bradley for fear of racism claims, but still went to vote for the other candidate. Leading to a widespread phenomenon where Bradley had a majority in the polls because everyone would, when asked, say they were voting for him, regardless of their actual intent.
We see the same here. I mean, one part is obviously media bias- the media is all too happy to fake the real numbers with tricks and loopholes, (or just lie) but its a similar effect. many people will say they support Hillary Clinton, or neither, because they fear the reprisal from peers and the media if they support Trump. They fear being called a racist bigot for declaring support for him because of how he's been demonized by the left. many fear being called a misogynist if they say they dont like Hillary.
This is much the same kind of effect as the Bradley Effect. Or, in fact, Nixon's Silent Majority, which was a similar instance of people considered to be overlooked by the media in favor of the vocal minority. a term that the Trump Campaign has declared the return of. And I would agree
it was studied that people were more likely to lie and say they were voting for Bradley for fear of racism claims, but still went to vote for the other candidate. Leading to a widespread phenomenon where Bradley had a majority in the polls because everyone would, when asked, say they were voting for him, regardless of their actual intent.
We see the same here. I mean, one part is obviously media bias- the media is all too happy to fake the real numbers with tricks and loopholes, (or just lie) but its a similar effect. many people will say they support Hillary Clinton, or neither, because they fear the reprisal from peers and the media if they support Trump. They fear being called a racist bigot for declaring support for him because of how he's been demonized by the left. many fear being called a misogynist if they say they dont like Hillary.
This is much the same kind of effect as the Bradley Effect. Or, in fact, Nixon's Silent Majority, which was a similar instance of people considered to be overlooked by the media in favor of the vocal minority. a term that the Trump Campaign has declared the return of. And I would agree
9
Nate, another difference is that USC asks about the panelists' intentions on election day while the rest ask some variant of "if the election were held today".
I suspect this might encourage some fence-sitting behavior.
They also allow for respondents to answer twice in any one given week through a methodological quirk of their sampling. I asked about this but the USC folks have reasonable justification for allowing double responses (doesn't affect the mean, etc).
I suspect this might encourage some fence-sitting behavior.
They also allow for respondents to answer twice in any one given week through a methodological quirk of their sampling. I asked about this but the USC folks have reasonable justification for allowing double responses (doesn't affect the mean, etc).
3
FWIW, it strikes me that there's an even larger problem with the USC/LAT poll. If it only contained just over 50% voters in the 2012 election, I suggest that it's far under-representing those voters. Newly enfranchised voters are largely those under 22 this year, who typically don't vote in very high numbers. And I suspect (albeit don't know for sure) that people who didn't vote in the last presidential election are not that likely to vote in the current one. And the corollary question is, how were the other 48% of the panel chosen?
But I agree with you in that this poll is very useful in tracking trending since it recontacts the same voters each time, and I agree that the shift is very significant (and backs up the shift in other polls).
But I agree with you in that this poll is very useful in tracking trending since it recontacts the same voters each time, and I agree that the shift is very significant (and backs up the shift in other polls).
8
Trump, the lying joke, the guy is nuts and always has been. He must hire people to build for him because after see and watching Trump he is really stupid and I don't think the guy could build a child's playhouse that came pre assembled. The only thing he is good at is spreading hate.
42
There are people voting republican because their daddy or grandfather did. It is like the scene out of Forest Gump where Lt. Dan says “there has been a member of my family that has died in every American war” but with the republican voters it would be “members of my family have voted republican in every American presidential season”. Interestingly enough, the results of both Lt. Dan’s family and the family members who vote republican, both pretty much end with the same results. The cult of “family” passing down through the years the dysfunction all held up proudly as if it was an honor to do so. Women and children being forced to vote republican because their dominating husband or father forced them to do so. Male driven religions, Adams rib, reducing women and children to be less than whatever man has been assigned to govern them. Meanwhile what the GOP stands for is against the very needs of those women and children and in doing so is actually against any healthy man’s family.
Liberals speak for the weak and oppressed; want change and justice, even at risk of chaos.
Conservatives speak for institutions and traditions; want order even at the cost to those at the bottom.
[Psychologist Jonathan Haidt studies the five moral values that form the basis of our political choices, whether we're left, right, or center]
Liberals speak for the weak and oppressed; want change and justice, even at risk of chaos.
Conservatives speak for institutions and traditions; want order even at the cost to those at the bottom.
[Psychologist Jonathan Haidt studies the five moral values that form the basis of our political choices, whether we're left, right, or center]
27
And there are a few strong examples like Katy Perry who decided to be liberalized and become a Democratic party member even though her parents are ardent Republicans.
12
By liberalized I believe you mean set free, allowed to have a mind of her own, be herself, make choices without her hyper religious parents, a true sense of strength to overcome such pressure.
12
Hummmm, I think your analysis accounts for a great deal of the staying power of the Republican Party. Inasmuch as it is based on emotion rather than reason, it does not augur well for the future of the country--or the world for that matter. The strain that these folks put on logic in order to perpetuate their "club" in power is astounding. Trump's economic policy, as announced today, is just more trickle down economics. It defies logic that such an economic theory would be proposed--especially now. The greatest share of income and the greatest share of wealth is already flowing to the small group of people who claim to be the "makers" and whose rising fortunes the theory predicts will raise all boats. Corporate accumulations of cash reserves are at record high levels and interest rates are at record low levels. If the rising tide has not raised ANY boats under the present circumstances, it strains credulity to think that more of the same would would produce any different result than it has in the past. If the "makers" haven't set in motion the kind of making that leads to the trickling down of prosperity under these circumstances what conceivable reason is there to believe that it ever would? It will only lead to a greater concentration of wealth along with the usual reductions in expenditures for essential programs. One of the definitions of insanity is said to be doing the same thing over and over while expecting a different result. Republicanism is bordering on insanity.
9
Imagine if you could "poll" the children of America...without their parents there to direct them how to respond.
I find children to be the best judge of character. They haven’t been polluted by society YET! Look at differences between how children and babies interact with President Obama or Hillary Clinton vs Donald Trump. Children have no problem showing how they feel.
Trump on his own child;
Trump on “The View” “I don’t think Ivanka would do that [pose for nude photographs] inside the magazine,” Trump says, speaking for his daughter. “Although she does have a very nice figure. I’ve said that if Ivanka weren’t my daughter, perhaps, I would be dating her.”
And then, as the words he had just spoken formed meaning in his mind, Trump asked: “Is that terrible?”
Joy Behar asks “Who are you, Woody Allen?” Everyone laughs, including Donald Trump, who remarks “That’s very good.”
“He told Rolling Stone that if he weren’t happily married and his daughter’s father, he’d — well, he didn’t finish the thought, but you can guess what he meant."
I find children to be the best judge of character. They haven’t been polluted by society YET! Look at differences between how children and babies interact with President Obama or Hillary Clinton vs Donald Trump. Children have no problem showing how they feel.
Trump on his own child;
Trump on “The View” “I don’t think Ivanka would do that [pose for nude photographs] inside the magazine,” Trump says, speaking for his daughter. “Although she does have a very nice figure. I’ve said that if Ivanka weren’t my daughter, perhaps, I would be dating her.”
And then, as the words he had just spoken formed meaning in his mind, Trump asked: “Is that terrible?”
Joy Behar asks “Who are you, Woody Allen?” Everyone laughs, including Donald Trump, who remarks “That’s very good.”
“He told Rolling Stone that if he weren’t happily married and his daughter’s father, he’d — well, he didn’t finish the thought, but you can guess what he meant."
22
Reuters have been tinkering with their model recently. They don't seem to have a convincing model. One thing is clear, as soon as all Bernie fans have started recognizing Hillary as their party candidate after DNC, Hillary's percentages have gone up enormously, while Trump's stays where it was or lower. This also means there is a big question mark. Is the Republican Party today considered too conservative and outdated for the Americans and needs a replacement? Time will tell.
17
Donald Trump is not a conservative.
4
"There’s a very good reason: People just don’t seem to report their past vote very accurately."
Really? Where's the evidence of this?
Really? Where's the evidence of this?
4
There is political science research to support the conclusion that a number of folks change their votes after an electoral outcome, leaning heavily toward the winner. People also overestimate in surveys having voted, as checked by getting their names, addresses, and checking the actual voter rolls. For example, in the data its fascinating how many people in 1976 said they didn't vote for Nixon, although they did based on his 1972 vote totals. Memory is a fickle thing.
14
Most of the polls are based on telephone calls made to registered voters. Many of Trump's supporters are not registered and may never even have voted before. I also hear that there is a shame factor which prevents many from acknowledging their preference for Trump. That's why we have secret ballots.
3
The USC/LA Times is a longitudinal panel survey, not a cross-sectional tracking poll like most other news outlet polls. It is completely valid, but is most useful for assessing attitudinal shifts, and can't be compared to other polling numbers. The design of panel surveys is prone to "panel conditioning", where the previous panel questions can bias subsequent responses. After multiple "waves" of questioning, the respondents are no longer spontaneous, but become more cautious and pay more attention media reports relevant to the questions. The details of the USC methodology to mitigate conditioning and seam (wave-to-wave) effects is unclear on their website, but in any case their results should not be lumped together with other "tracking polls. Apples and oranges!
14
Any poll that views Trump as "favorable" in any way, shape, or form should be seriously questioned, and have its demographic breakdowns explored.
14
Reuters has Trump within 2 points of Hillary as of 8/5/2016.
Poor Nate.
Poor Nate.
4
An average of all the polls had Clinton up by 6.8% on that same day so obviously some of them had her even higher.
10
It would be interesting to know the politics of the LAT reader base. One might conclude, given where it is located and the type of newspaper it is, that the panel is really not representative especially given the importance of the other party in California politics.
4
Comparison of poll results is a useless exercise, nor can you average different poll results. You have to control for the variance in the questions asked -- a near impossible task. Furthermore, responders are sometimes prone to jerking around the pollsters --saying the opposite of what they believe.
Witness the amazement at the results of the last British election. The USC poll provides continuity. The NY Times, WSJ, and Washington Post Polls are usually heavily cast with leading questions, which compromise the results. Suffice it to say, polling is in the eyes of of the beholder.
Witness the amazement at the results of the last British election. The USC poll provides continuity. The NY Times, WSJ, and Washington Post Polls are usually heavily cast with leading questions, which compromise the results. Suffice it to say, polling is in the eyes of of the beholder.
8
"Witness the amazement at the results of the last British election. "
Wrong. The polling averages all showed Brexit leading in the last two weeks. Only people who ignored polling averages were amazed. See
http://www.bloomberg.com/graphics/2016-brexit-watch/
and
http://www.huffingtonpost.com/entry/uk-brexit-polls_us_576a7fc8e4b065534...
Wrong. The polling averages all showed Brexit leading in the last two weeks. Only people who ignored polling averages were amazed. See
http://www.bloomberg.com/graphics/2016-brexit-watch/
and
http://www.huffingtonpost.com/entry/uk-brexit-polls_us_576a7fc8e4b065534...
18
It is too early to read polls, which is another way of saying polls are not very useful.
1
Frankly, the continuum of "bash Mr. Trump" via the NYT and ABC in particular notwithstanding other MSM media is to the point of ad nauseam! Every few minutes or so I google "Trump" and EVERY major newspaper and network promulgates a vitriol disdain for the gentleman. I do not ever recall a time when the top brass of a major party during an election year resigns and garners a few seconds in the medium as opposed to four minutes on a baby purportedly being tossed from a Trump rally which of course never actually happened. Then the audacity of those who were responsible for such a story NEVER capitulating doesn't bode well for their respective brand or integrity thereof, in my view. We look to the journalist for both sides of an issue then allow the reader and/or viewer to decide how it should be interpreted. To do otherwise as I was taught back in my Harvard days is nothing short of yellow journalism. I trust you understand my premise. Thank you in advance for the opportunity to opine. Cheers.
8
No. We look to journalists to report facts--and clearly delineated opinions--from which we can form our own opinions. If Trump said something and the Times, for instance, reports what he said and provides the audio & video (the baby incident comes immediately to mind since you said it never happened), it is not Trump-bashing (unless you are maintaining that Trump isn't, in effect, bashing himself). Statements he makes are news and are fact. You should read Jim Rutenberg's mediator article in yesterday's Times to get an intelligent picture of the conundrum journalists are in when covering Trump.
Question: if you don't get your news from what you call "mainstream media," where do you get it? And how?
Question: if you don't get your news from what you call "mainstream media," where do you get it? And how?
36
It's not yellow journalism. We live in perilous times and Trump's unfitness is a major story, with global ramifications.
32
You Google "Trump" every few minutes? Couldn't your Harvard-educated talents be put to better use?
52
Would appreciate a tweak and update to this column.
Any chance of your getting the raw, unweighted results and giving your readers a tool to do their own sample weighting by different factors?!
Could you also present data on the panel itself-- not just who the respondents said they voted for, but rather, their demographics, party affiliation, etc.? Why was this sample Trump heavy to begin with?!
Finally, how about raw data over time-- what have the percentages been for Trump for Clinton--and what are the Ns and the panel's attrition rates?!
It's all too easy for the panel to get smaller and smaller, as fewer people want to keep participating -- or, only those of particular preferences do.
In the interests of transparency and democracy all pollsters should provide the raw data and weighting tools to the public, the pundits, and the political analyses for their own determinations.
Any chance of your getting the raw, unweighted results and giving your readers a tool to do their own sample weighting by different factors?!
Could you also present data on the panel itself-- not just who the respondents said they voted for, but rather, their demographics, party affiliation, etc.? Why was this sample Trump heavy to begin with?!
Finally, how about raw data over time-- what have the percentages been for Trump for Clinton--and what are the Ns and the panel's attrition rates?!
It's all too easy for the panel to get smaller and smaller, as fewer people want to keep participating -- or, only those of particular preferences do.
In the interests of transparency and democracy all pollsters should provide the raw data and weighting tools to the public, the pundits, and the political analyses for their own determinations.
1
They don't give you the raw data. They started the poll July 4, and published the first numbers July 10 - the average of the previous seven days. Subsequently, they update it every day as a moving average, replacing the oldest number (8 days back) with the latest (i.e. the previous day's) and re-averaging. This is why the results change far more slowly than other polls.
You could deconstruct the moving average into daily snapshot numbers by making some assumptions about July 4 through July 10. Simple Excel spreadsheet.
Btw, the panel itself is pretty confusing - it is not clear from the USC website that it is the same 400 people each time or it is 400 out of a pool of 2000. If the latter, each day's population would have to be re-skewed (haven't heard that term much this election cycle)....all in all, if not defective, a very problematic methodology.
You could deconstruct the moving average into daily snapshot numbers by making some assumptions about July 4 through July 10. Simple Excel spreadsheet.
Btw, the panel itself is pretty confusing - it is not clear from the USC website that it is the same 400 people each time or it is 400 out of a pool of 2000. If the latter, each day's population would have to be re-skewed (haven't heard that term much this election cycle)....all in all, if not defective, a very problematic methodology.
4
When you have a poll that is constantly an extreme outlier compared to the other polls we should discard it from the list of acceptable polls regardless of what shortcomings we can identify analyzing the poll method. Rasmussen is another example that historically overestimated Republicans standings compared to the real elections results.
22
Indeed! Rasmussen is the reason Romney didn't have a concession speech prepared.
7
Hey, no one has ever asked me, feel abandoned! :) As an aside, I simply do not understand why the polls are not 80/20 in favor of Clinton, even with her problems, she is at least sane! We are truly sailing in uncharted waters, hope we do not get lost or drown!
49
If only the same scrutiny were applied to the polls favorable to Hillary. Why single out one that is favorable to Trump? They are all flawed. There is not one methodology that is the accepted standard.
1
They do and there is: try 538 for nuanced poll analysis.
19
Ashamed of voting for Trump. Then do not vote for him. But a lot of Republicans are bombastic like Trump and they are proud of showing that.
3
Well at least we know today if Trump says his lead in the polls is "huuuuge" - he really is lying!
12
Just another example of why I believe polling is, at best, useless and possibly a danger to the democratic process. Too many people will weigh their voting choice on the basis of who has the best chance of winning. They do not want to be tarred a loser by voting for an unpopular candidate, and that adds a momentum factor to an election that has no connection to the candidate's merits and flaws.
I, for one, see no upside to the constant polls and many potential downsides. I would like to see more attention paid to what the candidates think and less to what my neighbors think; I am voting for the former and not the latter.
I, for one, see no upside to the constant polls and many potential downsides. I would like to see more attention paid to what the candidates think and less to what my neighbors think; I am voting for the former and not the latter.
6
Quite a good point. Called herd behavior, which is also prevalent in stock trading.
3
I never lie about who I voted for because I do not worry about being on the losing side. It upsets me that my choice did not win but why lie about it? I had no idea so many voters lie about who they voted for or who they are going to vote for.
I voted for McGovern. I voted for Dukakis. I did not vote for Jimmy Carter when he won. I did not vote for Reagan. I did not vote for GH Bush when he won. I did not vote for GW Bush (the second) either time.
Ocassionally, I regret who I voted for. I reget not voting for Carter when he won. I ended up respecting and admiring him quite a bit. He was new to me when he first ran and I did not warm up to him until much later.
I am sure that I will not regret not voting for Trump even if he wins. He is not new. He has been on television and the news for decades. I have never liked nor respected him so I cannot vote for him. I will never warm up to him like I did with Carter and Gerald Ford as presidents.
I voted for McGovern. I voted for Dukakis. I did not vote for Jimmy Carter when he won. I did not vote for Reagan. I did not vote for GH Bush when he won. I did not vote for GW Bush (the second) either time.
Ocassionally, I regret who I voted for. I reget not voting for Carter when he won. I ended up respecting and admiring him quite a bit. He was new to me when he first ran and I did not warm up to him until much later.
I am sure that I will not regret not voting for Trump even if he wins. He is not new. He has been on television and the news for decades. I have never liked nor respected him so I cannot vote for him. I will never warm up to him like I did with Carter and Gerald Ford as presidents.
4
I teach sampling methodology and I still found this discussion confusing. These discussions are really useful in getting people to think about how polls work and educate people on statistics and sampling more generally. It's also a nice opportunity in this case to talk about the difference between longitudinal and cross-sectional data. I had hoped to use this piece to illustrate those points for my class, but it doesn't do a good enough job.
One key point - not mentioned here - is that the composition of the sample is even more critical for longitudinal than cross-sectional polls because of dropout issues and because you've just got one shot at getting it right. This critique of the decision to weight by past voting behavior is spot on. An equally important question (not addressed) was why they decided to weight this way and not a probabilistic stratified random sample of past voters plus new voters and people who didn't vote. The latter category is particularly interesting because I don't know how many past non-voters will come out this time - I'm sure someone has that data.
So these are great points. But I think they're writing to us survey nerds and making a broader opportunity to explain things better by writing better.
One key point - not mentioned here - is that the composition of the sample is even more critical for longitudinal than cross-sectional polls because of dropout issues and because you've just got one shot at getting it right. This critique of the decision to weight by past voting behavior is spot on. An equally important question (not addressed) was why they decided to weight this way and not a probabilistic stratified random sample of past voters plus new voters and people who didn't vote. The latter category is particularly interesting because I don't know how many past non-voters will come out this time - I'm sure someone has that data.
So these are great points. But I think they're writing to us survey nerds and making a broader opportunity to explain things better by writing better.
34
Nellie, if you don't already, the published methodologies of 538 and PEC could be used in your teaching. These two sites pretty-much ignore each other as rivals (posting something non-negative about 538 at PEC never survives moderation), but both use some really interesting techniques worthy of study. Both leave me behind in some of the statistical details, but they converge as the election nears. Both have stellar prediction histories.
As of today, these are their predictions for electoral votes and odds of winning:
538: 361-177 86%-14%
PEC: 341-197 85%-15%
As of today, these are their predictions for electoral votes and odds of winning:
538: 361-177 86%-14%
PEC: 341-197 85%-15%
6
This is tantamount to weighting by partisanship and the conditions of the last election.
I thought we had this debate in 2008 and decided this was a Very Bad Idea....
I thought we had this debate in 2008 and decided this was a Very Bad Idea....
8
Translation - This LA poll methodology favors Trump/GOP voting by weighting method, but still shows a 7 point slippage for Trump from its previous results.
57
What I don't get about the USC/Dornslife poll is how it gets weighted by poll aggregators like this site and fivethirtyeight. It has a release almost every day, but this is just to report the trailing average of the ongoing survey. It's not like they've got a new independent poll every day. I would think that the natural way to address this in a model would be to just replace the USC/Dornslife number every time it comes out, instead of counting it along with previous releases of the same poll in some kind of weighted fashion. But at least when I look at 538, I see several versions of this poll covering different time periods included in the model. That seems to overweight this poll, but maybe I'm missing something.
4
538 puts most of the weight on the most recent USC/Dornslife poll (0.77 in the current forecast), with small weights for the prior polls (e.g., 0.03 for the prior day's poll)
http://projects.fivethirtyeight.com/2016-election-forecast/national-polls/
http://projects.fivethirtyeight.com/2016-election-forecast/national-polls/
1
The 538 model also incorporates changes from one polling occasion to another in polls run by the same organization - the "trendline," as they call it.
1
Many of the pollsters are now using the probability models based on Bayes' Rule. So they continuously contacts random members from the population and update the probabilities.