Presidential Race 2016

[inactive_user]

The Drumpfinator works freaking perfectly. My Internet experience has really improved.

[Juan_Raigada]

So there’s this new CNN Poll out that has Trump at 49% nationally in the GOP, which is pretty terrible.

However, looking at the methodology, I’m forced to think that this poll is pretty much garbage.

Now, they say that they weight the results by national demographics, which is fine… but when you look at the actual demographics of those who were polled, you see that they would effectively prohibit weighting for major demographic groups, because they contained NONE of them.

For instance, every single person polled on the republican side was white… but more importantly, EVERY SINGLE PERSON POLLED WAS OVER 50 YEARS OLD. Seriously, they polled ZERO people who were under 50. Now, I’m not a statistician, but doesn’t this prohibit this poll from accounting for anyone under 50 years old, in any possible way? It basically assumes that they have no impact at all.

Further, those polled were 100% from the south, and 100% from suburbia.

So… that’s pretty much a joke, right?

i think what the poll says is that there are not good numbers for those segments under N/A, maybe because they refused to answer some questions or because the numbers where so tiny that percentages without adjustment can’t be extrapolated. If you look at page 32 and 33 the totals don’t add up unless you assume some non-south, non-white respondents.

Not that the poll doesn’t seem to have a bias, mind you. It’s just not as egregious as you paint it.

[magnet]

when you look at the actual demographics of those who were polled, you see that they would effectively prohibit weighting for major demographic groups, because they contained NONE of them.

For instance, every single person polled on the republican side was white… but more importantly, EVERY SINGLE PERSON POLLED WAS OVER 50 YEARS OLD. Seriously, they polled ZERO people who were under 50.

Not exactly.

Some subgroups represent too small a share of the national population to produce crosstabs with an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error larger than +/-8.5 percentage points are not displayed and instead are denoted with “NA”.

[Timex]

i think what the poll says is that there are not good numbers for those segments under N/A, maybe because they refused to answer some questions or because the numbers where so tiny that percentages without adjustment can’t be extrapolated. If you look at page 32 and 33 the totals don’t add up unless you assume some non-south, non-white respondents.

Not that the poll doesn’t seem to have a bias, mind you. It’s just not as egregious as you paint it.

What parts of the poll are you looking at that suggest non-southern, non-white, non-old respondants? I’m not seeing what you’re talking about on pages 32 and 33.

[Timex]

Not exactly.

Ok, so potentially they polled some people under the age of 50… maybe. And if they did, it was so few as to not be able to generate a reasonable margin of error.

So yeah, seems like it’s a pretty worthless poll unless you’re interested in what old southerners living in the suburbs think.

[Telefrog]

Does any of this matter? Trump is still going to win the GOP nomination.

[Juan_Raigada]

For example, in Page 32 it gives Ted Cruz a total of 14% of respondents saying he’s the most trustworthy, while it also says 12% of white respondents said so. If white respondents where at 12% but the total is at 14% it implies the existence of non-white respondents modifying the final number.

I don’t know why I picked up those two pages, they were the most obvious at a first glance?

[Timex]

This is John Oliver’s piece on Donald Trump, which is (obviously) awesome.

[magnet]

So yeah, seems like it’s a pretty worthless poll unless you’re interested in what old southerners living in the suburbs think.

Not exactly.

They polled people across the US, with responses in the 35-49 age group on up. Look at the cross-tabs for Q2 and Q3.

The registered Republicans were predominantly old southerners in the suburbs.

[Timex]

Not exactly.

They polled people across the US, with responses in the 35-49 age group on up. Look at the cross-tabs for Q2 and Q3.

The registered Republicans were predominantly old southerners in the suburbs.

But that’s the part of the poll they are focusing on, saying that trump has nearly half of the Republican base.

But their view of the Republican base is heavily skewed.

[Murbella]

Making fun of Trump’s hands has a long history, and apparently gets under his skin bad.

Also, lulz at saying “trump hasn’t sunk to that level” about anything.

This is totally the facial expression that tons of these people have at his rallies. They are absolutely fanatical, like girls at a boyband concert, or Hitler’s followers.

Everyone should be fighting to destroy Trump at this point. You can hate the other republican candidates all you want, but put that shit under your hat for the time being. Save it until Trump is done, THEN attack whoever the republicans put up.

I don’t know, i guess i just find this sort of thing juvenile, more so than making fun of someone’s height or other Trump comments. YMMV though.

Really though, Cruz is bat shit crazy as usual, but Marco Rubio IS trump now. The scene from star wars where the emperor is encouraging luke to kill vader and take his place comes to mind.

[magnet]

But that’s the part of the poll they are focusing on, saying that trump has nearly half of the Republican base.

But their view of the Republican base is heavily skewed.

How do you know that it’s skewed? If anything, the CNN data are telling you that old southerners in the suburbs are the base.

But honestly, I don’t think you should read that much into the crosstabs. Their reporting cutoff was MoE of 8.5%, and a lot of the subgroup numbers come pretty close to that.

It’s entirely possible to have an sample in which NONE of the subgroup numbers are worth reporting, i.e. every column is filled with “NA”, yet the full sample analysis is still worthwhile. It just means that the study isn’t powered for a subgroup breakdown. As a thought experiment, imagine polling 5 people in each of 50 states. You would never want to report results by state-by-state, but the full sample of 250 could still be significant.

[Dan_Theman]

It may or may not be skewed, but a sampling error of 8.5% is toward the high side and so not reporting anything beyond it is sensible. I can also understand the decision to cut it off, because the longer you run it the less yield you get per call, the more expensive it is to run, and the less valid it can be if you’re trying to measure the impact of a certain event, like a debate (other events may occur with an individual that would influence their answers after the measured moment, and the longer it takes to get their answer the greater chance that will occur). Still, they have to find a way to up the n next time.

[magnet]

I think the sample size is fine for the purpose of the poll, which to examine the nationwide support for Trump among registered Republicans. He polled at 49%, the next closest was 16%, and the MoE was 5%. That’s enough to conclude that he is doing really well.

The poll wasn’t meant to address his regional support, so you can’t really argue that the regional MoEs are a methodological defect.

[triggercut]

You kinda need to read Josh Marshall’s opinion piece at TPM regarding a whole lot of reaping, sowing, and poultry that’s come home to roost.

For pure prose stylings alone (“unchurched libertine”, “Taj MaWall”) it’s pretty genius. But he also brings the point home that this is the logical outgrowth of debt ceiling fights, government shutdowns, Planned Parenthood baby-part merchants, and wishful thinking goofballery regarding immigration reform and Obamacare. The GOP over the last 8 years especially built up a debt of paranoia and hate…and now this has happened.

There’s some metaphor or analogy here about a hostile takeover, though hostile takeovers don’t usually take place because of excess debt. They’re proxy battles or stock purchases. But there are numerous ways that profligate spending and excess debt can leave a highly leveraged company vulnerable to a guileful schemer who strangles the ownership and takes the carcass for himself. Some version of that is the story of Trump - a raid on a hopelessly leveraged GOP ‘establishment’ which barely realized that it scarcely exists.

[Oghier]

I saw that piece over at TPM, and I thought it was outstanding. It also makes me wonder how long it would take the GOP to return to some kind of rational course. How can they bring that part of their base back to reality? Will they even try?

[inactive_user]

That’s a great editorial. The technical debt metaphor is quite apt.

[triggercut]

Hey Chris Christie, I hope the check at least cleared.

This was particularly cringe-worthy:

STEPHANOPOULOS: – finally, you talk about Donald Trump as the person who can be trusted. He still maintains that he saw thousands of people celebrating in New Jersey after the 9/11 attacks. You said that didn’t happen, it was not factual.

So if he continues to hold that position, isn’t that a slander of your state citizens?

CHRISTIE: Now, listen, I don’t think he means it as that at all, as a slander of my state’s citizens. You, what I think, though, is when I say he’s – can be trusted, this is a guy who, when he makes promises, he keeps them. Um, and I’ve seen that over the course of a 14-year relationship with him.

[Dan_Theman]

I think the sample size is fine for the purpose of the poll, which to examine the nationwide support for Trump among registered Republicans. He polled at 49%, the next closest was 16%, and the MoE was 5%. That’s enough to conclude that he is doing really well.

The poll wasn’t meant to address his regional support, so you can’t really argue that the regional MoEs are a methodological defect.

If you’re going to break it down into demographics that don’t have enough representation, then yeah, you can. The rest of the poll is fine.

[Telefrog]

Hey Chris Christie, I hope the check at least cleared.

What the Hell?

(BEGIN VIDEO CLIP)

CHRISTIE: I tell everybody who goes to a Donald Trump event, if he gets asked a question, just ask him how. First, he says he’s going to build a wall across the entire border between the United States and Mexico.

How?

How is he going to make the Mexicans pay for the wall?

How?

They are a sovereign nation.

(END VIDEO CLIP)

STEPHANOPOULOS: You’re backing him now.

What’s the answer?

CHRISTIE: The answer is he will do it –

STEPHANOPOULOS: How?

CHRISTIE: – the fact – the fact is that he’s going to have to answer that question. And he will.