Election post-mortem
While Biden has won a historic victory in the popular vote with over 80 million votes, the highest in American history, as well as a winning margin greater than any American president since FDR, the polls predicted that he would do much better. So this result, along with the one in 2016, will only reinforce the idea that "the polls don't mean anything" for people who don't delve into them more deeply.
What seems to have happened with the national vote is that while Biden's support was as predicted, most of the undecided votes went for Trump. So it was clearly a mistake to assume that the undecided voters would automatically be split between the two candidates. In the future, undecided voters should be added to the margin of error, meaning if you have one candidate at 48% and the other at 46%, with 6% undecided, you shouldn't assume that the final result will be 51% to 49%. It could also be 52% to 48% in the other candidate's favor.
Also, some states were spot-on, for example, Georgia was polled at a close tie with Biden slightly ahead, which is exactly how the election turned out. Ann Seltzer also got Iowa correct. Seltzer specializes in Iowa, so the future of polling may be for pollsters to focus on a specific state rather than attempt to poll the entire country. For example, the results in Ohio and Florida were way off.
The other issue is the effect of the demonstrations after George Floyd's murder. Ruy Teixeira, a fellow at the Center for American Progress, has discussed this, but I came up with it before reading his article, so I'm not just repeating what he said. The Republican propaganda campaign linking the Democrats to the demonstrations was very effective, and cut into Biden's lead in suburban areas. If the demonstrations hadn't happened, Biden would have won by a much larger margin. Based on the example of Richard Nixon in 1968 and 1972, people seem to turn toward conservative Republicans in response to social unrest, and that apparently happened this year, too, although not in numbers high enough to give Trump a second term.
The theory that Democrats do better when turnout is higher was pretty much relegated to the dustbin. This election had higher turnout than any other one since 1908, yet the Democrats didn't do anywhere near as well as expected. Both parties have been working under the assumption that high turnout helps Democrats, with Democrats seeking to register more voters and make voting easier, while Republicans do the opposite. Apparently these strategies don't have an effect, so we may see fewer attempts by Republicans to suppress votes in the future.
The next question will be what happens in 2022. The Democrats did very well in 2018, unusual for a midterm (unless, of course, turnout doesn't matter anymore). The party in the White House usually loses seats in Congress in the first midterm. However, it's possible that Trump motivated more people to vote (either for him or against him), so if he's not on the ballot, that will affect the results.
What seems to have happened with the national vote is that while Biden's support was as predicted, most of the undecided votes went for Trump. So it was clearly a mistake to assume that the undecided voters would automatically be split between the two candidates. In the future, undecided voters should be added to the margin of error, meaning if you have one candidate at 48% and the other at 46%, with 6% undecided, you shouldn't assume that the final result will be 51% to 49%. It could also be 52% to 48% in the other candidate's favor.
Also, some states were spot-on, for example, Georgia was polled at a close tie with Biden slightly ahead, which is exactly how the election turned out. Ann Seltzer also got Iowa correct. Seltzer specializes in Iowa, so the future of polling may be for pollsters to focus on a specific state rather than attempt to poll the entire country. For example, the results in Ohio and Florida were way off.
The other issue is the effect of the demonstrations after George Floyd's murder. Ruy Teixeira, a fellow at the Center for American Progress, has discussed this, but I came up with it before reading his article, so I'm not just repeating what he said. The Republican propaganda campaign linking the Democrats to the demonstrations was very effective, and cut into Biden's lead in suburban areas. If the demonstrations hadn't happened, Biden would have won by a much larger margin. Based on the example of Richard Nixon in 1968 and 1972, people seem to turn toward conservative Republicans in response to social unrest, and that apparently happened this year, too, although not in numbers high enough to give Trump a second term.
The theory that Democrats do better when turnout is higher was pretty much relegated to the dustbin. This election had higher turnout than any other one since 1908, yet the Democrats didn't do anywhere near as well as expected. Both parties have been working under the assumption that high turnout helps Democrats, with Democrats seeking to register more voters and make voting easier, while Republicans do the opposite. Apparently these strategies don't have an effect, so we may see fewer attempts by Republicans to suppress votes in the future.
The next question will be what happens in 2022. The Democrats did very well in 2018, unusual for a midterm (unless, of course, turnout doesn't matter anymore). The party in the White House usually loses seats in Congress in the first midterm. However, it's possible that Trump motivated more people to vote (either for him or against him), so if he's not on the ballot, that will affect the results.