1. As Goes Washington, So Goes the Nation? As early as 2010, I’d noticed that Washington state’s primary predicts November national outcomes fairly well. The idea is this: Washington has a primary where all the candidates run on a single ballot, with the top two advancing to the general election. The specifics have changed over time – in particular, the primary is held in August now rather than in September – but regardless, that unique format gives us a bit of a dry run as to how voting will go in the fall (California has a similar system, but its primary is held much earlier)…
The data for 2020 was genuinely surprising. I had expected that it would look like 2018, when Democrats won about 62% of the vote overall in the September primary. Instead, the Democrats’ performance this year was slightly worse than it was in 2016: They won about 55.2% of the vote in 2020 versus 56% in 2016.
While this also doesn’t look like the Democratic performances in the very good Republican years of 2010 (50.1% of the vote) and 2014 (51.9%), Republicans don’t need a large popular vote victory like they achieved in those years (the state of Washington has also become more Democratic overall since then).
If we use a regression analysis to predict vote shares in the fall based upon performances in the summer, the predictions are virtually identical to the predictions for 2016: Democrats are projected to perform on average 0.8% better than they did in 2016. Using the median (to account for potential outliers), they are projected to perform 0.3% better. Given “error margins” on regression predictions, that is effectively a projection of no change.