And finally, the US House of Representatives. There are 435 of these to predict, so I won't waste time. Tabulated data is provided in a supplementary post for those who wish to play with it for yourselves.
A number of these races have polling. I have used the most recent, likely voter polls from 2020 in these races and multiplied the results by a factor so that Republican + Democrat = 100%, then represented as the Republican share (thus Arizona 1's result of 46.20% indicates 53.80% support for the Democratic candidate). Races where candidates are uncontested or at least have no opposition from the other major party are also factored in here, but marked by party name rather than as a polling value.
There are also plenty of seats that are foregone conclusions so no-one wastes resources conducting polling there. I've decided on a very safe margin where seats are called when the 2018 vote is at least 60:40 in favour of the incumbent:
This fills in a lot of gaps, probably as a result of rampant Gerrymandering in a number of states.
In a third pass, we approximate a swing and apply it to outstanding districts. The swing is calculated for a district as 2020 polling - 2018 result. Therefore in Arizona 6, the swing is 47.67% - 55.20% = -7.53 (or a 7.53 percentage point swing away from the Republicans).
These swings have been averaged for the states as a whole, then applied to remaining districts as an approximation of the localised swing. In the example of Arizona 6 above, this is the only assessable district, so the Arizona swing is assumed at -7.53 below. The other averages are:
Arkansas: -3.22; California: 4.11; Colorado: -4.73; Florida: -0.09; Georgia: -1.48; Illinois: 9.44; Indiana: -10.82; Michigan: -10.09; Minnesota: -0.07; New Jersey: 3.96; New York: -0.97; North Carolina: -6.10; Ohio: -3.04; Pennsylvania: 0.69; South Carolina: 1.79; Texas: -2.25; Virginia: -4.38; Washington: -1.66; and Wisconsin: -0.66.
Arkansas 2 has polling split 50:50 and was a solid Dem seat until a sudden flip in 2010. I'm going to take a gamble and predict a D Gain
Connecticut 5 has been blue since '07 and isn't likely to flip after this divisive campaign. D Hold
Missouri 2 has only been blue once since 1987. Clinton couldn't flip it. Obama couldn't flip it. I doubt Biden will do better. R Hold
Nevada 2 has always been Red. R Hold
Nevada 3 and Nevada 4 are contested but went blue when Trump was elected and stayed blue in the midterms. With Nevada as a whole putting Biden ahead, D Hold is the safe bet
Oregon 2 is a pretty clear R Hold
Oregon 4 and Oregon 5 are obvious D Holds
West VIrginia 2 and West Virginia 3 are safe Republican districts. R Hold
This gives the final prediction:
