Biden’s superiority is the most stable on record

This year, 95% of all individual polls so far have shown results in 6 points from the average. Basically that’s what you expect if you do a lot of polls and the race is not moving (ie. The only change is the statistical noise from sampling). This is a very small range historically.

The previous low for a 95% confidence interval built similar was 8 points (2012). The median cycle displays an interval of 95% from 13 points from the poll average. In other words, about double the range of polls that we have seen so far in 2020.

Sometimes, of course, the range of results can be wider than the median cycle. Lyndon Johnson had a gain ranging from 35 points to more than 60 points above Barry Goldwater in the early months of 1964. Jimmy Carter opened 1980 with a 30 point advantage or greater in several polls over Ronald Reagan. In early May, his lead fell to a single digit, and he even trailed Reagan in several polls.

Indeed, there seems to be some relationship between the initial poll movement and how well the initial poll predicts the outcome. In cycles with smaller initial movements, the average difference between the polls for four months and the final result is 5 points. In cycles with larger than normal initial movements, the average difference between polling during that time and the final result is 14 points.

In other words, it will match the previous pattern if the current poll is generally close to the result. This is what you expected, given that more voters have a stronger view of Trump than any other incumbent since at least 1980 (when the question was first asked), as I noted earlier.

The obvious catch here is that the 5 point difference between polling stations and November results can be huge this year. Biden rises on average only 6 nationally. A 5 point shift towards Trump would mean a 1 point victory by Biden in the general election. This can greatly leave him on the wrong side of popular voting and college elections. Remember, Trump lost 2 points nationally in 2016 and is still attracting victories in college elections.

And we are only talking about averages here. The 1956 campaign, which showed a limited shift at the start, made Democrat Adlai Stevenson do 10 points better than the initial vote that pegged it.

The point is that while the stability in the 2020 race has so far been projecting progress towards minimal movement from now on, that does not in any way guarantee Biden wins. There are still various possible outcomes.