Overtime Politics Thread (WARNING: Possible fraud)

[Adam Griffin]

I haven't looked at the poll to even see if there's crosstabs, but I do know that my 2012 state-by-state breakdowns for Obama's votes by race showed TX as essentially 33/33/33 Black/White/Latino. If you favor them a bit more toward blacks and whites and Sanders is somehow doing equally well among (non-Hispanic) whites and latinos, then it's possible.

[Adam Griffin]

THEY'RE LITERALLY PULLING STUFF OUT OF THEIR ASSES

I'm not sure if anybody else has suggested it thus far, but I've been convinced from the beginning that "they" are really just one person who is literally making figures up, and not actually polling anyone. My suspicions were heightened when it became so easy for them to go from "well we're just a small-time group and we don't have time to ask all of these questions" to "OK we're going to do it".

Now, this Michigan poll...

Supposedly, they sampled 405 people. Now, let's look at the 2012 exit polls. In Michigan, 3% of the voters were non-black, non-white, non-latino voters. Among the Democratic electorate, this group would comprise at most 5% of the electorate, but in a primary...likely less. However, let's just assume that for argument's sake, it's 5%.

In their poll, Asian voters were 45-44; Nat-Am and Other were 44-42. Five percent of 405 people = 20 people. This means at minimum, there should be 5 percentage points' difference between the two candidates in said group if just one person more supported Clinton than Sanders, or vice-versa. However, these are two separate groups, splitting (what should be no more than) five percentage. Let's say the two groups were sampled equally and combined, comprised five percentage. That means that just one more person supporting a given candidate would give you a ten point difference.

Yet in these "crosstabs", there is only a 1-point difference among Asians and a 2-point difference among others. You'd have to have 12% of the poll's respondents be Asian in order to have such a scenario be possible; 24% of the poll's respondents be Nat-Am/Other for the same. Even if you assume "rounding error", you'll still looking at these two groups needing to comprise roughly 25% of the people surveyed.

IT'S BOGUS, MADE-UP BS - PURGE ANY OF THESE FROM THE DATABASE IF THEY EXIST

[Adam Griffin]

I think the Republican poll might be legit. My guess is that they polled 26 Hispanics, about 12 Asians, and 8 "Native American/Other"s, judging from the percentage numbers.

Out of the 396 polled, that suggests 2% Nat/Other, 3% Asian, and 6.6% Hispanic, which doesn't really sound too far from Michigan demographics.

For a Republican primary, though, it's way off. Hell, even for the general election, it's way off, as the notion that both Latinos and Asians alike do not comprise anywhere nearly the same share of the electorate as they do of the population (for Latinos, it's about 40%; for Asians, about 50%). In reality, the R primary poll should be anywhere from 96-98% white in Michigan.

@Adam Griffin

I emailed the site about what you had mentioned, and I got this response:

I double checked the spreadsheet and it does appear as though there is an error in one of the formulas on the Democratic side. The Republican poll doesn't have the error. I'll be posting the updated information shortly.

Thank you for bringing it to my attention.

- Fred

And see, this is just another one of those "oops - let me quickly fix/change that" moments. I'm still wholly convinced he's just plopping his own numbers down in a spreadsheet.

[Adam Griffin]

Wait, we're still discussing this "firm" as if it's a real thing? I already showed how they're literally just making up numbers on page 15. Stop discussing this "pollster" as if it's anything other than one bored guy making up numbers and claiming that he has a staff, in the hopes that he'll get a paid client at some point in the future. This thread should be locked and any future posts about it should be deleted.

Real polling costs a lot of money: if he was actually conducting as many polls as he was posting, then the only possible explanation would be that he's a very wasteful multi-millionaire with nothing to do. If he was a very wasteful multi-millionaire with nothing to do, then he'd actually hire the staff necessary to perform real polling with real methodology, real questions and real results. Ever wonder why well-established and respected polling firms aren't cranking out multiple polls of irrelevant states each and every week? Yeah...

[Adam Griffin]

Overtime Politics. Is not. A Real. Pollster. Repeat after me.

It's not a matter of even being a bad pollster. He is not a real pollster at all. It doesn't matter that the one dude who is literally sitting in his basement typing random figures into spreadsheets guessed halfway accurately because he looked at the mean of polling.

[Adam Griffin]

Guys, I just wanted you all to be the first to know that I have started my own polling firm, Griffin Analytics™, and my team of 4 people officially started dialing random numbers in Georgia today. I'll be spending approximately $1,000 of my own money each week for the remainder of 2016 so that I can supply the American people with a true snapshot of the 2016 election. We'll actually be dialing voters based on ZIP codes, which I believe to be a superior way in which to hyper-localize the actual electorate and produce superior results.

We'll be calling Georgians through Saturday and will publish our first poll's results Monday morning.

[Adam Griffin]

Wait, we're still discussing this "firm" as if it's a real thing? I already showed how they're literally just making up numbers on page 15. Stop discussing this "pollster" as if it's anything other than one bored guy making up numbers and claiming that he has a staff, in the hopes that he'll get a paid client at some point in the future. This thread should be locked and any future posts about it should be deleted.

Real polling costs a lot of money: if he was actually conducting as many polls as he was posting, then the only possible explanation would be that he's a very wasteful multi-millionaire with nothing to do. If he was a very wasteful multi-millionaire with nothing to do, then he'd actually hire the staff necessary to perform real polling with real methodology, real questions and real results. Ever wonder why well-established and respected polling firms aren't cranking out multiple polls of irrelevant states each and every week? Yeah...