I have spent a lot of time this past week thinking about polls. Reputable pollsters always report their methods and techniques and disclose any problems that they see in their own methodologies. But unlike the results that we see in scientific and engineering journals, our ability to authenticate the results of the poll and redo the experiment never takes place. There is no reconciliation of results and that alone makes all polls unscientific. This is never mentioned in those sting polls or by the pollsters themselves.
Last week Rasmussen reported that Donald Trump had a 28% approval rating and a poll done the same week by the Pew Foundation—an organization that I am very familiar with, stated that his approval rating was 51%. A smaller poll done last Monday by a news organization reported Mr. Trump’s approval to be at 50%. How can these results be so different? There is methodology available to test polling data based on known patterns of answers in known groups with known voting patterns.
The key question is the assumption of the inputs. If 49% of people voted for Trump in November and 50% for Biden, why would anyone use a population of subjects with 35% Trump voters and 50% Biden voters? In local precinct polling, the exact percentage from last election data can be used and a precinct-by-precinct model can be built for larger groups in State and Federal elections. One other not mentioned but too common problem with this methodology is that percentages and absolute values are mixed up. When percentages of a precinct with many voters are mixed in with percentages of precincts of very few voters and the math is based on changes in percentages, the results of such a poll are worthless.
I will disclose upfront that I have conducted an unscientific poll by any mathematical standard but a poll that easily conforms to the Rasmussen/Pew/WSJ-CBS-any poll that Karl Rove’s organization has lately run/ CBS Wall street Journal poll standard. I say this not to be mean spirited, but I can’t find a defined stated standard in any of these polls and I have no ability to validate any poll—mine included, unless I can rerun the poll on the same subject. Wouldn’t it be nice to line up all the pollsters and have them poll the same 1500 subjects and because their questions are all different see how different the poll results would be?
I polled 1500 people in Canyon, Boundary, and Bonner Counties because these counties represent all corners of our state. I didn’t include Ada or Blaine because they are in the middle of the state and I was looking for “diversity” and as any statistician knows diversity is never found in the middle of anywhere. President Trump received an 85% approval rating. Mr. Biden got 17%—the extra two people—a guy and a gale(there I go getting absolute values and percentages mixed up again) we counted anyway knowing that they were Canadian citizens, but they were thinking about immigrating to this country if Biden loosens up on the rules about dual citizenship.
Just like the voting machines in Wayne County Michigan after 9PM election night I didn’t really count votes I just extrapolated and added a percentage to the Trump vote taking that same percentage away from the Biden vote. Who knows what really happened in Wayne County, but I am sure it was close to what happened in my poll? For skeptics, I kept all my data on 4 pages in my note pad. I spilled some pizza and coke on the pad on my way back from Sand Point, but I bet we could get the FBI to reconstruct all the data. My data has been at least archived and the Jimmy Carter Foundation for voter fraud has been offered the data for their review—more than we can say about the Nov. 3rd election. Archiving votes and reconciliation of votes should be part of the process.
I added Nancy Pelosi to my poll, and she got a 7% approval rating and Chuck Schumer got 4%. NBC/CBS/MSNBC/CNN/ABC/FOXNEWS/Washington Post /New York Times/ Idaho Statesman all got less than 2% approval. Even if Mr. Trump was polling at 28%, he is at least 4 times better than Nancy and seven times better than Chuck. But our pollsters did something unique and experimental. They showed pictures of the various politicians and media celebrities. It was interesting that with a confidence limit of 0.00123 most people polled couldn’t tell the difference between Nancy Pelosi and Peggy Noonan. They also couldn’t tell the difference between Adam Schiff or John Kasich, Mitt Romney, or Paul Ryan. They could only guess at who Hunter Biden was but thought that he had done business of some sort with Eric Sowell in China. They struggled trying to tell the difference between Mitch McConnell and Casey Stengle.
The reason wrote today in a satirical fashion is twofold. I wanted to point out the fact that polls are not scientific. In less one can recreate the experiment and reconcile results it is not science—despite what the pollsters will tell you.
Secondly, there have been hours of YOUTUBE video from MIT and STANFORD and even more articles in newspapers about the results in key swing states being manipulated by algorithms but people far brighter than I—MIT and CAL TECH mathematicians missed the point of mixing up absolute values and percentages. A great man once told me—do the math before you do the calculus. The same man told me that it doesn’t matter if you do the math or the calculous—just do it right. Finally, he told me that when geniuses make mistakes, they usually make simple mistakes because they always check the complicated stuff. I am no genius and I have sure screwed up a lot of complicated and simple stuff. Just because somebody is a PhD elite scientist or mathematician, or pollster doesn’t mean they can’t make mistakes. The same thing can be said about experts as what they say about “Harvard Men”—You can always tell an expert—you just can’t tell him much.”—ask Jake Tapper or Mike Wallace or Rachel Maddow.
Finally—when the people who voted for Trump in my poll were asked if they would travel 500 miles and sit in the cold for 6 hours to see Mr. Trump at a rally 100% of them said yes. When those who disapproved of Mr. Trump were asked the same question about our new President and Vice president they answered—”will we get free stuff?”
There is nobody else in the country including Garth Brooks or LeBron James and especially Nancy Pelosi and her legions of syncopates who could draw a crowd like Donald Trump. Not Mr. Biden or Ms. Harris. Not Mr. or Mrs. Obama. They know this. They know we know this. His connection with 75million people frightens the Libs and the RINOS to their bones. There is no poll that can put a number on loyalty or commitment to country and cause and most of America understands that. WE do—they still don’t.
One reply on “About Numbers and Polls”
Dr. Livingston is right. Polls are rarely truly accurate. It is a rare question that can be answered either ‘yes’ or ‘no’ and that is what many polls are predicated upon.