My Authors
Read all threads
4 August 2020 #BetweenTheLinesDotVote Analysis

Averages polling and Tilted Models...Again...

Yes, here we go again. Complete lies from the polling world, worse than I've ever seen. Start with this article:

politico.com/news/2020/08/0…
2) And would you look at this headline and sub-headline!

‘Hating Joe Biden doesn’t juice up their base’: Key swing state slips away from Trump

Trump has trailed in every public poll in Pennsylvania since June.
3) Before diving in, look at this quote:

"Joe Biden has an overall early lead in the state of 6 percentage points, according to RealClearPolitics’ polling average, and has led Trump in all 12 public polls released since the beginning of June."
4) So, Trump fans, its all over. Better start selling your red hats on eBay now, before they're worthless, right. Pennsylvania is gone, and with it, the election. That's what this article wants you to believe. I'll come back to that.
5) And you Democrats, starting laying in your champagne supply and practice your bottle uncorking stylings, get ready to win big!
6) Not so fast, bukos on either side of the line. The flaws here are too many to list. Let's begin with the entire idea behind Real Clear Politics and their averages of polls, so beloved of the talking class on TV. Surely, when we average the polls we'll get a real clear picture.
7) I have no beef with the creators or operators of this site, I don't. But I 100% disagree with their methodology. Had I bought into in 2016, my analysis would have been as faulty as everyone else out there, predicting for HRC. The idea of an average, in polling is dumb.
8) As we'll see yet again below, each poll must be judged on its own merits. And the great majority of them fail on easy to identify parameters. As I continuously say:

It's The Model, Stupid!

We must always look for the chosen model to determine if the poll is valid or bent.
9) Again and again, I dive into the provided data, and find the models to be flawed. Easily identified. Easily disproved. Obvious leaning. Thumb on the scale. Done by polling scientists who know what they're doing, and they do it for a purpose, an agenda.
10) Remember, we're talking about a single poll, one at a time. It is easy to find the tilt, poll after poll. So, what happens when you average out polls that all employ the same deceitful methods? You get a flawed average, but this average has MORE credibility.
11) The reasoning is simple, but deceitful. If we take an average of polls, then, OBVIOUSLY, the mistakes in one are corrected by another and the average must REALLY indicate whats ACTUALLY going on. Right? It's an average. How can it be wrong?
12) If you'd like to see how well this method works, go check out this page, showing where all the polls stood at the end of the 2016 campaign. I'll show you the key image below.

realclearpolitics.com/epolls/2016/pr…
13) As you can see, by averaging it all together, RCP only managed to be 100% wrong, with a 3.2 % advantage to Clinton. Please note, they'd have had it even worse if they hadn't included the LA Times Dornsife poll, the ONLY one calling for Trump.
14) I hope you're following. With a single exception, every one of the polls they averaged got the call wrong, in varying degrees of distance, but in a binary call, you're either 100% right or 100% wrong. The average was to precisely zero benefit. Maybe it was even detrimental.
15) Okay, I don't mean maybe. It is detrimental. You cannot improve bad data by pooling it. And, when you do, and claim the high ground of presenting an even more credible, and even more wrong average, you do additional harm.
16) Mind, I am NOT judging the RCP people, themselves. They may be true believers in what they're doing. I'm only saying that it is demonstrably wrong, and I completely disagree to the point of rejecting their method with prejudice.
17) It was by precisely that reasoning, and that method of individual poll by individual poll analysis, that I selected the LA Times poll, in 2016, as the ONE AND ONLY source of data I analyzed. I also always said, my analysis can never be better than my data informs.
18) I chose the very best poll, and I did by the very methods I'm sharing here. And I'm saying we can do even better. Obviously, you'll be hearing more about that as we roll.

Now let's turn to state level polling.
19) In 2016, and up until very recently, I never paid the slightest attention to polling at state level. I do not yet have a feel for it, but that's rapidly changing. There is one thing that transfers directly, though, from national polling to state polling: The Model.
20) What exactly is a Model? It's one's guess, first about who will actually vote, and secondly, how they identify, and then last, who they'll vote for. Please remember, it is a guess, no more. And, there's no getting around it. To poll is to use a model. A predictive model.
21) At the highest level, we're just attempting to imagine how many Democrats, Independents, and Republicans will show up and vote. Then, we're looking to see how loyal the Ds and Rs are to their guy. Any variation there can turn the entire election.
22) In today's politics, however, it is the Independents, and every data I see indicates they're growing, taking from both formal parties, who's break tells the real story. How Independents lean appears to be, growingly, the actual factor that matters. The difference that counts.
23) Purely at the instinctive level, the model I lean toward, personally, is 30% D, 40% I, 30% R. I do NOT believe that more Ds vote than Rs. I also don't, however, believe in any of the data informing what has taken place so far. I do NOT buy the data I've seen on point.
24) In my own polling efforts, just now getting rolling, I plan - for the moment - to simply use three 33%s for the model I will employ: 33% each for Ds, Is, and Rs. If I find a way to improve that model I will, and I suspect I will. I will always post my model BOLDLY.
25) Getting back to the connection between national and state level polling, the same question of the model remains. That one pivotal question never changes.

What
Is
The
Model?

Always ask that one question, to determine what you buy or reject in each and every poll.
26) Getting back to our Politico article above, they graciously allowed us access to a single poll. From the esteemed team at FOX. Yes, sadly, I say that sarcastically. I have NO ESTEEM for the FOX polling operation, and even less so today.

static.foxnews.com/foxnews.com/co…
27) Thankfully, the PDF above is only 11 pages. I am so sick to death of the 300 or more page PDFs I have to arduously scroll through to find the model, somewhere, hidden deeply within its bowels. Today's was not so difficult to find. What we seek is on page 3.
28) Remember the headline? Trump has lost Pennsylvania, and we know this by an AVERAGE of polls, as reported by RCP. We already know, here at the beginning of August, Trump has lost Pennsylvania. And the FOX poll is one of those included in the average.
29) And would you look at that model!

D = 48%
R = 41%
I = a whopping 11%

Uh huh. Yeah. I'm impressed...NOT!!!
30) Small and inadequate as my own data still is, it indicates a roughly 4 - 1 break for Trump among Independents. I do NOT have state level information, yet. So, that's just a general number, and NOT properly founded on sufficient data. But, it's what I've got, so far.
31) But let's be generous. Let's call for a 50/50 split among Independents. You can do the math. If you upped the number of Independents in the poll, taking equally from Ds and Rs, you'd have a VERY different picture from the poll, wouldn't you?
32) And then, what about the 48% Ds to 41% Rs? Okay, Pennsylvania MIGHT break that way. It's not the worst such I've seen. Remember, there are two D strongholds, Philly and Pittsburgh. Do you really think they outweigh the rest of the state to that degree, today? I'm not sure.
33) I'm on the limb here, my Democrat friends. I don't think you have quite that large an edge in Pennsylvania. But, if you only focus on party identification, I can run with it. What I can't do, is give Biden the enthusiasm edge. I don't see it.
34) So, if I'm a Democrat, I'm NOT TAKING ANYTHING for granted in Pennsylvania from this poll. NOT ONE THING. And then, what if Pennsylvanian Independents are A) a bit greater than 11%, and B) break for Trump, not Biden?

Warning Will Robinson. Danger. Danger.
35) If you need a coffee break, class, now would be a good moment. We have to dive down an additional layer deeper, and it won't be easy going. I trust you, though. You're up to it. But seriously, if you need to take a break, do so now. We'll wait for you to get back.
36) Duly caffeinated? Excellent. We have some media work to do now. A first principle is this. Never reject a story due to the story teller. Judge the story on its own face. Even the greatest liar may, every once in a while, be telling the truth. Maybe even against his own will!
37) A corollary of that principle is this. Always seek out information from those you disagree with the most. Essentially, I personally HATE Politico. But, the fact that Drudge - and I don't like him much anymore - led me to their story today is a good thing!
38) And follow the bouncing balls!

Drudge
Politico
Real Clear Politics
FOX Polling

Isn't that an interesting path? That path alone is instructive.
39) If you believed what you hear, you'd think that FOX was Trump's Office of Propaganda. Any person on the MAGA side knows this is simply 100% false. When I watch FOX I see about a 50/50 split for and against Trump.
40) But when I follow FOX polling, I see no less than a 90% tilt AGAINST Trump. This matters. And it doesn't just matter to MAGA people. It matters to Democrats. FOX should serve you every bit as much as it serves anyone else. It owes you the truth. Not an agenda based poll.
41) There is, in all this, another hand. If you know a storyteller has an agenda. If you know a storyteller to be a proven liar. Then, you are certainly, and rightly, allowed your cynicism when listening to that storyteller's stories. Politico is a known and proven liar.
42) FOX polling is NOT a known liar, yet. They have managed to hold the sanctified position of TRUSTWORTHINESS. They do NOT deserve this, yet they absolutely enjoy it. The model they employed above, for Pennsylvania, should cement their deceitfulness.
43) Let's go back to Drudge, now. Why this headline, from this source, including these averages of the polling, including this one poll, itself? If you follow those bouncing balls, you can derive the agenda. The story falls on its own lack of merit. What's the agenda?
44) Helicopter with me up to a higher level. Check this out! We have another story, from Politico, about Drudge. This is very interesting. Look at this:

politico.com/story/2013/01/…
45) 17 January, 1998. That's just over 22 years ago. On that day, Drudge began to reincarnate Woodward and Bernstein in taking down a sitting POTUS. What could be a greater dream for a journalist? The Drudge outfit is working its butt off to do the same again today.
46) While every story, and every poll, must be judged individually, on its own merit or lack, there is the perpetual question of motivation. And when it comes to Drudge today, there can be no question. His outfit's goal is to take down Trump.
47) Allow me to back peddle, a bit. I don't thing the current operators at Drudge actually care about outcomes. They only care about clicks, advertising revenues, profit. They see the profit coming from antipathy to Trump, so there they go. Every story is anti-Trump.
48) Do NOT let me get started on ethical capitalism. I have a book, or more than one book to write on that topic. Just hear me. I 100% REJECT unethical capitalism. We will not, we must not, discuss communism or socialism now. Please help me with my focus discipline.
49) When you look at Politico's motivation for this story, picked up by Drudge, and including FOX's poll, you see a deep level agenda. Again, a liar may tell the truth. But, hearing a known liar's story, you are allowed to distrust the storyteller's motive.
50) Trump may or may not lose Pennsylvania in November. The false polling reported gives us less than no real information. It is agenda driven, and the motivation matters. Honest polling is needed. If you wish the truth, help me procure it. Go here:

BetweenTheLines.Vote
51) We ask the zip code of each respondent. We have NOT yet built our real data chops. We struggle forward each day, with our site, our data, our analysis of the sparse data we've yet procured. But, I assert to you, we're honest. We seek ONLY the truth.
52) Yesterday, I discovered the tipping point on our data. We need roughly 220 people to answer our poll each day. New or old is not important. We encourage previously polled souls to respond again each day. Just once per day.
53) Be it new or old, what our true goal is, is 70 - 75 respondents from each of the 3 categories: D, I, or R.

We look at that goal as a minimum. The model, as you might have already seen is:

D = 33%
I = 33%
R = 33%
54) How many zip codes are there in Pennsylvania? If we had adequate respondents, identifying from those zip codes, we would be able to give you honest polling for all of Pennsylvania. Zip codes are the answer. All the zip codes in a given state. Make sense?
55) That is the answer to the question: how will Pennsylvania vote in November? Who will win its Electoral Votes? Proper polling, in advance of the election, would call that outcome correctly. I don't know that my new outfit will be able. But we're hellbent on getting this right.
Thread ends at #55.
Missing some Tweet in this thread? You can try to force a refresh.

Keep Current with Pasquale "Pat" Scopelliti

Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!