What the polls mean – The Ledger

The Ledger

Today, I got a tattoo. Its a little dolphin, cresting an invisible wave, moving forward. Why I decided to get one, after decades of turning my nose up at such things, would take more ink to explain than that which actually went into my arm, and would mean a ton less; suffice to say that a year ago, I would have placed the probability of my only slightly painful visit to Phil Snyders studio down at the Black Swan (as anything more than a spectator) at less than 5%. How things turn, right?

The possibility of infection from such a procedure is less than 10%, if you do as you are told somewhat more if you ignore safety caveats. Probabilities rule our lives we measure the likelihood of something happening against something not happening again and again during the election season.And, like running out and getting a tattoo, we often choose to read the tea leaves of polling in ways that suit our view of what the outcome should be, rather than accepting what the numbers tell us will (likely) be.

FiveThirtyEight, one of my favorite polling sites, has recently posted the odds of Donald Trump winning this presidential election at 12 in 100 (the chances of Joe Biden winning the race is 87 in 100). They arrive at these probabilities through the use of algorithmic models, employing data from a wide range of reliable polling figures, and add to the data every time a poll is posted. Without going into great detail about how these models work, they do seem to work pretty well.

This does not mean that Donald Trump will lose. It means that after running their models 40,000 times, the probability of a scenario playing out in which he would win is fairly tiny. Two things separate the polling results this year from results in the past: first, the models are more robust, careful and focus far more on scenarios of the Electoral College than in the past, and second probably far more importantly is that almost 30% of the people who are expected to vote in this election already HAVE voted.

Why does this matter to pollsters? Arent they trying to predict the eventual outcome which, as we well know from 2016, will not be known until (perhaps late) on election day? They are.But the by-word, since scientific polling began, is if the election were held today, this would be the probability of X." This meant that the ebb and flow of public opinion was simply a projection, week by week, of what people would probably do on election day itself. Thus, a candidate might be doing remarkably well in September, but fall off or even collapse in a heap before the first Tuesday in November.

Things are just different in 2020. Since early voting and mail-in voting have been taking place for weeks, the election is, in fact, being held today." Or yesterday. Or for some voters, has been over for weeks. Which means as we get progressively closer to the election, the probability that the probabilities expressed in the models and the polls they are based on become progressively more reliable.

Voting taking place during the week that the poll is conducted can be expected (with normal skew from who is voting and where) to be closer to the polling results. These votes cannot be called back. The certainty factor the probability (how certain are you) grows with each passing day in which votes are cast.

By the results so far, time for the Trump folks to catch up is simply running out.Biden appears to be winning. Which means that if you have not yet voted, whichever side of the presidential equation you are on, your vote is worth exponentially more with each passing day.

Models aside, the race is not yet over, but it's getting down to decision time.


R. Bruce Anderson (randerson2@flsouthern.edu) is the Dr. Sarah D. and L. Kirk McKay Jr. Endowed Chair in American History, Government, and Civics at Florida Southern College in Lakeland.

Go here to see the original:
What the polls mean - The Ledger