With ten days away until the Election Day, 2020, more than enough polling data has come forth to forecast a President Trump re-election on November 3, 2020.
National Polls are 100% B.S.
Back in 2016, I shocked the world by telling the world that national polls would be wrong (see my work: WHY EVERY PRESIDENTIAL ELECTION POLL FOR 2016 YOU SHALL SEE OR HEAR WILL BE WRONG). I did so again in 2020 (see my work: Well It is Election Time, Again, and Fake News USA Today Is Touting an Unscientific National Poll Claiming Trump is Behind, Again. When Will They Ever Learn?).
As a longtime data analytics professional with an graduate-level academic background in public opinion polling, I know that national polls for US presidential elections are unscientific and thus no results can be projected to the universe of voters.
State Polls Suffer from Serious Errors
Most do not realize that state polls suffer from significant flaws if taken at face value, as presented. Polling firms get it wrong because they fail to pull from a random sample of likely voters from the way populations are distributed in respective states. In short, pollsters take too much sample from mega urban areas and ignore polling residents in rural and suburban areas.
When pollsters do this, they merely are surveying urban voters and particularly, registered Democrats. A study of cities with populations of 75,000 or more will show that all are mostly occupied by voters registered as Democrats.
Yet, many states, the suburban / rural population taken in total is a significant percentage of total population.
So while you can take state polls to mean how voters in bigger cities will vote, you can not trust those results for statewide voting.
Joe Biden's True Opponent for 2020
The 2016 True Dollar Journal Forecast Model Predicted the Trump Victory
The True Dollar Journal Forecast for 2020
So how did I generate this map?
The model assumes pollsters almost exclusively survey urban voters. The model assumes the 2020 rural vote will be the same as the 2016 rural vote.
So I break out the vote totals in states as 75,000 or less as non-urban (which I call rural for easier means) and 75001 and up as urban. Then I calculate an average of the latest state polls published at Real Clear Politics over the last several months. Using the average for each state, I apply the percentages for Trump and Biden to the 2016 urban vote tally for each state.
In summary,
- Rural people will vote in 2020 the same as in 2016
- The 2020 state polls are really urban polls for the respective states
- The percentages from the 2020 state polls apply to the 2016 urban votes cast
If I am right that pollsters are randomly grabbing respondents mostly, if not exclusively from urban areas, an increase of urban voters for President Trump, even though he loses by the polls, actually is a win for President Trump when comparing Biden's take relative to Hillary's 2016 take.
Surprising Results
Updates
State Polling Done Right
In an accurate survey for any state, a pollster would need to create uniform population groups and then take at random from each group and complete an equal number of questionnaires for each group to get the total completed questionnaire counts. For a margin of error of plus or minus 5% at the 95th confidence level, that means 376 completed questionnaires.
Polling firms do not do this because the costs incurred to get hold of rural and suburban voters is much higher.
Having worked in this business of public opinion polling as a director both for a marketing research firm and for a newspaper, I can tell you that getting the questionnaires done by deadline becomes the goal even knowing that randomness rules need to be broken. Many firms cheat their clients. When the deadline presses, much like fishing, many simply go to where respondents are jumping and the task to complete questionnaires is easiest.
But doing so borks the data. In short, for many polls, you can not trust the results as anything more than opinions, which can not get projected to the universe of voters.