One America News Georgia Poll Is Bad, Should Feel Bad
How to make a bad poll in just 486 easy steps!
With Fox News and Rasmussen both showing Donald Trump down by double digits, the president has taken to scream-tweeting about "Fake Suppression Polls," unlike his "VERY GOOD internal Polling Numbers," which he won't be sharing with you, so don't even ask. Luckily, Trump's favorite off-brand Fox News substitute, the One America News Network (OANN), has cooked up some polls in swing states to please the sad MAGA man.
Faced with the daunting task of showing Trump winning bigly, OANN has started polling swing states like North Carolina, Arizona, and most recently Georgia. You already know these polls are junk . But let's reach up the ass end of that Georgia poll and see exactly what went wrong.
First Order of Business: Check the Survey Method
According to OANN, "The survey was collected from respondents using interactive voice responses and an online panel of cell phone users." What the hell does that mean? A good pollster will aim for 50 percent landline and 50 percent cell phone respondents, usually live callers only. Presumably, the Gravis polling company conducted its survey by calling and texting people, but we have no way of knowing. Was it cellphones? Landlines? Text? Twitter poll? That sentence of gibberish could mean a lot of different things.
Second Order of Business: Sample Size
Any good poll, will sample about 800-1500 people, depending on the state. The OANN poll sampled 513 likely voters, off to a great start.
Third Order: Weighting
All pollsters weight their results for demographics because the cohort of people willing to answer the phone and donate ten minutes of their time to a random stranger is likely not representative of the electorate generally. But this is OANN, the "news" people who took time out of a White House press briefing to ask if Chinese food was racist . If there's fuckery here, it's probably in how they weighted the results. And what a coincidence , they didn't tell us how they weighted the demographics. Could their baseline be 2016? It's probably not 2018, since the GOP got blown out then and GOP Georgia Governor Brian Kemp almost lost. (Probably did lose, if everybody had been able to vote and have their votes counted, but that's another post.)
Okay, let's check out the actual numbers and see if we can see what they did here.
Crosstabs: Let's Talk About Race
Let's give Gravis credit here, because its top line result is very reasonable. If Trump won Georgia by three, no one would be surprised. But let's look a little closer at these horseshit crosstabs.
Okay, sure, it's possible Trump could get 12 percent of the Black vote in Georgia. He got nine percent last time and is doing better nationally than in 2016 with African Americans. That said, 12 percent seems implausible, particularly in light of the ongoing protests against police violence.
WTF Is Going On With the Asian-American Vote?
According to OANN, 21.6 percent of Asians prefer Trump, 20.6 percent like Biden, and almost a whopping 58 percent are uncertain. WTF? You have seven percent of African Americans, 19 percent of Hispanics, and six percent of whites undecided, but only 41 percent of Asian Americans have made up their minds? Come on!
Nationally, Asian-American voters are becoming increasingly Democratic. Two-thirds of Asian-Americans supported Clinton in 2016, and 77 percent backed down-ballot Democrats in 2016. This ain't 1992 — the Republican candidate won't win Asian Americans.
Trump Making Big Inroads With Hispanics, You Bet!
According to OANN, 36 percent of Latinos back Trump for re-election, an 11 point increase since 2016. Donald Trump has made some progress with Hispanic voters nationally, so perhaps he could get to 32 or maybe 33 percent in Georgia. But 36 percent is unrealistic, as is the 19 percent who are undecided. Joe Biden, who doesn't spend half his day shitting on immigrants, will probably win at least 60 percent of the Latino vote.
But When It Comes to White People ...
I have no issues with the white vote here. Seems like a very good representation of where Georgia's melanin-challenged population is at right now. Good job, OANN!
Age Is Just a Number, But It's a Pretty Important One
What the hell is going on with that 18-29 number? According to OANN, Trump has 18 percent support with young voters, Biden is at 54, and a full 28 percent are undecided. Trump is at 31 percent with young people nationally, but can't crack 20 percent in Republican-leaning Georgia? Meanwhile, Clinton took 63 percent of millennials in 2016, but OANN wants us to believe Biden is doing nine points worse? That's just straight bullshit.
The 30-49 numbers are no better. Trump 57, and Biden 35? What the fuck? OANN says they weighted for voter demographics, but unless they weighted based on the 30-44 crowd in Wyoming, these results don't make any sense. In 2016, this vote went for Hillary 51-44 in Georgia. Apparently there has been a fundamental shift in this crowd between 2016 and 2020 and only OANN noticed it.
And speaking of fundamental shifts in the electorate, if Biden really is winning the 50-64 crowd by 10 points as OANN predicts, then Trump is in deep shit. For comparison, Fox has Trump leading by 11 points with Georgia voters over 45. In 2016, he won this cohort 57-41, and in 2018 Brian Kemp's margin was 57-42. If OANN's numbers were right, Trump would lose Georgia in a landslide.
The senior vote number isn't too bad, though. It looks like a reasonable snapshot of where Georgia seniors are right now. OANN gets its second gold star!
Gender
The splits by gender are horseshit, too. In 2016 Trump was at 60 percent with men in Georgia, and 43 percent with women. According to OANN, those numbers are at 54 and 42 percent today, meaning Trump became eight points less popular with men and one point less popular with women. Leave aside that suburban women have been abandoning the GOP in droves. There is no way in hell Trump can carry Georgia by three percent if his margin with men is down to just 15 points. That would probably lead to a Biden victory in the four to five point range.
And so on!
There aren't enough pixels in the world to go through every goofy thing about this poll, so here are some of the highlights:
GOP Senators David Perdue and Kelly Loeffler are getting 20 percent of the black vote;
Trump has a 25 percent approval rating among African Americans, 51 percent approval rating among Asian-Americans, and a 62 percent approval among Hispanics;
Brian Kemp has a 48 percent approval rating among African Americans, 71 percent approval rating among Asian-Americans (with none strongly disapproving), and a 77 percent approval rating among Hispanics;
Kelly Loeffler has a 40 percent approval among African Americans;
Asked about voting behavior in 2016, 78 percent of African-American respondents in this poll voted for Hillary Clinton, as opposed to 11 percent who voted for Trump (she won them 89-9);
Hispanic participants in this poll? 31-30 for Trump in 2016 (actual results were 67-27 Hillary in 2016);
In 2016, 16 percent of Hispanic respondents in this poll supported Gary Johnson, and 14 percent supported Jill Stein. (In the actual election, six percent supported Gary Johnson, and there wasn't enough data on Jill Stein's vote);
In 2016, white respondents to their poll say they backed Trump 65-25. (they backed him 75-21);
There is no question about Trump's handling of coronavirus or race relations;
There is no crosstab asking about religion. (I want to know what Georgian Jews think of Trump.)
Polling is never an exact science, and it's really hard. But Gravis Marketing and OANN disrespect the profession when they put out shit like this.
But we guess their job is just to make Trump feel better, so ...
[ OANN Georgia Poll / CNN Exit Polls 2016 ]
Follow me on Twitter at @TheDataGoon !
Click here to support your Wonkette.
I don't have a pencil sharpener. Actually, I do, but why bother?
Yeah, before more American lives were lost than all armed conflict since and including Korea. Let's not count that time.