USC Dornsife/L.A. Times Daybreak National Tracking: 11/7 - Trump +3.2

[cinyc]

USC Dornsife/L.A. Times Daybreak Poll
Trump 43%
Clinton 40%

Weighted Panel of 3,000 Voters; MoE +/- 3.
http://www.latimes.com/politics/la-na-pol-usc-lat-tracking-poll-20160705-snap-story.html

This is new daily tracking poll, which debuted today.  Some info on the methodology:

You must be logged in to read this quote.

It appears that they ask "Someone Else" instead of naming Johnson/Stein.  I can't find the numbers for "someone else" or the undecideds in the linked article.

[Buh her emails!]

Who?

[cinyc]

Who?

University of Southern California Dornsife College (their Liberal Arts division) and the Los Angeles Times, L.A.'s biggest newspaper.

[JRP1994]

Not sure about USC Dornsife, but LA Times polls have an A- rating by 538

[AOC Stan]

So is this going to be our thread for the daily poll?

[cinyc]

More on the methodology here:
https://uasdata.usc.edu/data/election-poll

They basically ask 1/7th of the panel who they are going to vote for and the percentage chance they will vote per day, so 450 or so people are requestioned every day.  Everyone theoretically gets asked once a week.  They then weigh the results based on likelihood of vote and demographics to come up with a new topline at midnight (presumably Pacific time) every day, using the previous week's results.  

Friday's results were closer than Thursday's, probably 42-41 Trump, but I can't say for sure because USC didn't provide real numbers, just a graph.

[Mehmentum]

I guess this means Clinton will lose the election.

[Xing]

Now I guess Trump will win the PV easily, while losing the EC by more than Romney did.

/s

[IceSpear]

Wow, after being Safe R two days ago, it now looks as if the election is Safe D again.

Ratings change: Safe D  --> Safe R

Actually, it would be Safe R if Trump led by 1 point. The fact that he leads by 3 makes it Super Ultra Mega Rock Solid R.

[Ebsy]

LA times literally admits that this is the first poll from the pollster. Total junk.

[Seriously?]

Wow, after being Safe R two days ago, it now looks as if the election is Safe D again.

Ratings change: Safe D  --> Safe R

Actually, it would be Safe R if Trump led by 1 point. The fact that he leads by 3 makes it Super Ultra Mega Rock Solid R.

Come on man, let's not overreact to a few polls. I wonder if Trump will win the popular vote by 3 but lose in a landslide in the Electoral College, though? The NBC polls definitely give me cause for concern! Maybe the election is still Likely D, after all?

Ratings change: Safe R --> Likely D 

Okay, I'll stop.

I am pretty sure we're in this different tranche of how they handle the likely voter or even registered voter electorate right now that is explaining the difference in the polling. Those clinging to a D+5 or D+6 model have Clinton with a lead. Those that are letting the electorate stay where it is based on the sample they get have Trump leading. More data points need to be made available before I can confirm.

[john cage bubblegum]

This is like the RAND/American Life Panel tracking poll from 2012 that surveyed the same group of people.  Even if the topline is questionable, it could be good for tracking trends. 

I wonder if RAND itself is going to be doing it again this cycle.

[Ebsy]

So this actually isn't really a poll and should probably not be recorded as such.

[AOC Stan]

LA times literally admits that this is the first poll from the pollster. Total junk.

actually, https://alpdata.rand.org/index.php?page=election2012 The site of the USC polling indicates that this is the exact same team which developed RAND in 2012 and correctly predicted the 2012 election based on their own polling + panelists, so I'd wait to jump to calling something junk just because it doesn't show your candidate ahead...

[AOC Stan]

This is like the RAND/American Life Panel tracking poll from 2012 that surveyed the same group of people.  Even if the topline is questionable, it could be good for tracking trends. 

I wonder if RAND itself is going to be doing it again this cycle.

same team that developed that, developed this: https://uasdata.usc.edu/data/election-poll

[Maxwell]

Questionable methodology.

[cinyc]

So this actually isn't really a poll and should probably not be recorded as such.

No.  It's a real tracking poll called a panel survey that uses the same panel over and over again instead of finding a new random sample every day.  The panel is supposedly a representative sample of the electorate (with the pollsters even providing tablets and internet access to those without a computer).   There is no magic in recruiting a new panel every day if the original sample is representative - it just costs the pollster more money.

Panel surveys have been used for tracking polls in the past.  They're probably better for spotting trends than the toplines alone.

The other quirk that this pollster uses is weighing responses based on probability of voting instead of using a traditional likely voter screen.

[Ebsy]

lol the crosstabs are ridiculous junk as well. We'll see if this s[Inks] tracker improves at all as time goes on, but for now it is firmly in the junk category. The only interesting part for me is that the voters expect Clinton to win by a wide margin.

[Seriously?]

lol the crosstabs are ridiculous junk as well. We'll see if this s[Inks] tracker improves at all as time goes on, but for now it is firmly in the junk category. The only interesting part for me is that the voters expect Clinton to win by a wide margin.

I don't think that's a good narrative for Clinton. As this race stays closer than expected and Trump looks like he can pull it off, the undecideds may think it's OK to get aboard the Trump train.

It's usually the candidate that beats the expectation game that gets the boost in from the undecideds in the end.

[pbrower2a]

Dubious methodology. Questionable randomness.

[cinyc]

Today's numbers:
Trump 42.2%
Clinton 41.8%
(951 respondents)

Yesterday was:
Trump 42.1%
Clinton 40.9%
(1193 respondents)

http://www.latimes.com/politics/

[nfrstan12]

That's a good trend. It went from Trump +1.2 to Trump +0.4. If this continues, then Clinton will win every vote on Election Day.

[AOC Stan]

New numbers:
Trump - 42.8
Clinton - 41.7
https://uasdata.usc.edu/data/election-poll
based on 1,828 respondents

[cinyc]

7/17 Numbers:
Trump 42.9%
Clinton 42.2%

[Donnie]

I am not sure about this tracking poll, but L.A Times had always
very accurate polls in the past.