How reliable is the La times/usc poll ??

[agnostic102]

Im usually on twitter arguing with right wingers. As i feel like if you cant contribute by canvassing then the next best step is to donate and win hearts and minds online or atleast fight back agianst the attacks on hillary. Anyway was arguing with this guy and he kept claiming how the latimes daily tracking poll is showing a huge trump lead when it was tied just a week ago or so. I told him it was a bullshit poll and its not reliable and i could show him a dozen polls that show opposite.

The issue is though he claimed it was a model that the rand corporation came up with and they did it in 2012 and it was one of the most accurate polls in that election. Is this true? anyone familiar with this poll agency or model? i dont want to post the link to the poll encase it is some bs poll but wanted to get some thoughts on it by people who follow the polls and nate silver more closely then i do.

[onecaliberal]

1. Ask president Romney.

[MFM008]

6. Lol

Exactly.

[Cicada]

2. New method, at least one obvious W

They weight their sample to have "correct" numbers of those who voted for Obama and Romney in 2012. But we know from other polls that too many people in polls say they voted for the winner. So some of the "Obama voters" in their sample actually voted for Romney. So that is a small republican bias. And the recent surge for Trump comes partly from 4% afro-American supporting Trump a week ago to 20% supporting him now. That sounds implausible to me. It is a novel untested polling method so we need to be wary. Also the polling group is headed by Dan Schnor
, a Republican operative. He was Gov Pete Wilson's press secretary, the governor who won running ads of Mexicans streaming into California like cockroach es. School thinks Repubs need to woo Hispanics with appealing immigration reform but maybe he tilts republican in polling choices.

[agnostic102]

4. Thank you

Cicada! Thank you so much as soon as i saw headed by a republican i was like thats all i need to know. With regards to trump getting 20 percent of the black vote. I spent sometime looking over there poll and analyzing YES there was a jump in trump getting the AA vote from 4 percent to almost 18 percent.

BUT heres the kicker.. literally a few times he dropped RIGHT back down to 4 percent. That cannot happen.. you cant go up 400 percent in one day polling then down 400 percent in next. NOTHING happend in between for it to cause something that dramatic. So the jump that high was bullshit.

That being said im starting to see polls at least in invididual states like north carolina and ohio with trump having about 15-20 percent of the AA vote. I wont speak to these polls specifically but even polls show young voters not nearly as supportive as hillary as they were with obama.

What worries me about this election is not the polls so much. Its the indifference to voting in general. In midterm elections we keep getting our out butts kicked because the young, the AA vote the millennials just dont turn out. We CAN NOT afford that during this election. I know that hillary is putting a drive to appeal to the millennials and i hope that works. Basically we need a get out to vote with everything being on the line.

[Bernardo de La Paz]

3. Reliable per movement (tracking), unreliable for absolute values. . . . nt

[RAFisher]

5. The results are skewed but the 1st derivative (slope/change) is pretty good

The poll doesn't re-sample. It's net even a secret that this is the short coming of this method. However it does seem to be correct in showing if Trump or Clinton is gaining or losing ground.

[BzaDem]

7. Their model has plusses and minuses.

On the plus side, variation week over week cannot be due sampling error, because they sample the same group each week. This makes it a somewhat useful guide to trends.

On the minus side, the absolute support for the candidates could indeed be affected by sampling error, because they only picked one sample that they keep polling week after week. This likely causes a bias, though the direction is unknown. (As another poster indicated though, they seem to weight according to which people voted for which candidates in 2012. This likely causes a bias in favor of Republicans, since some people who didn't vote for the winner nevertheless say they did. This effect has been reproduced over and over.)

Also, asking the same people over and over again could actually affect the poll results. People who are repeatedly asked about candidate preferences might be more likely to follow the latest news about the campaign (either at a shallow level or deeper level). It is unknown what direction this might bias the result in, but it could indeed bias the results. This is in contrast to most polls, which mostly call people who have not previously been polled (because most pollsters call people at random, and the chances of any particular person being polled is small).