I am continuing a discussion that was in the NYT/Siena poll thread, to stop that thread from going off topic. Mods can merge that over to here if they want.
An assumption is that Dems are maxed out in uber-safe D seats so gains will disproportionately be in moderate seats.
Dems are close to maxed out in uber-safe D seats for getting voters to swing to them, because there are few remaining R voters who could possibly that is true. However, here is a list of Congressional districts ordered from highest to lowest in terms of swing from Obama '12 margin to Clinton '16 margin, except instead of showing the swing in terms of % of the vote, it is converted to the probability of an individual voter swinging. This is a different thing because vote shares are mathematically capped at 100% (and have a floor of 0%). This is a different way to look at swing as compared to what is shown on the Atlas maps, but is in some ways more illuminating. It is basically the swing you are used to looking at, but converted onto a logistic curve (with the districts that swung to Trump being the same thing, except opposite/backwards). Notice that this is basically exactly a list of the most highly educated white/liberal districts, also with heavily Hispanic districts mixed in:
NY-12, for example, is basically an uber-version of GA-06, TX-32, etc. Although those districts have a bigger swing in terms of % of the vote, in terms of individual voters there was more swing in NY-12, VA-08, etc. That fact is just ordinarily masked by the fact that there are not many Republicans in those places in the first place.
Anyway, the point here is that although there is indeed less swing in the heavily Dem districts, that is because of the way you are used to looking at the data. And the greater probability of voters swinging in these heavily Dem districts partially offsets the fact that there are fewer potential Republicans to swing there.
And just because it is interesting, these are the ones on the other end of the spectrum - the most particularly "Trumpy"/least Clinton (in comparison to Romney/Obama) districts. By this measure, OH-06 is the "most Trumpy" Congressional district:
(I could also post the other districts in between if there is interest)
Additionally - and more importantly - Dems are not maxed out in
turnout in uber-safe D seats. Other things being equal, Dems are more likely to gain additional votes from a turnout increase in the most heavily Dem districts than in less Dem districts, because more of the non-voters are also Dems. In other words, in absolute numbers, there are more white liberal college educated non-voters per district who could be turned out in districts such as NY-12 or VA-08 than per district in districts such as GA-06 or TX-07. So this is an important reason why Dems are not actually close to capped out in terms of the popular vote they should expect from Uber-Dem districts - and should actually probably expect a disproportionate share of any anti-Trump white college educated surge to occur in those uber-dem white college educated districts like VA-08 (which may help Kaine in the Senate race that he is already going to win, but is not going to flip any additional house seats).
This same list I posted above is, I think, not too bad of an index of districts where Dems are likely to gain the most from turnout. With turnout, it is also important to understand how additional votes translate into the % in the national popular vote as opposed to how they translate into the % of vote in the individual district. Here's an example with 2 simplified districts to illustrate how this, each with 100 voters:
Suppose you have 2 districts:
- District A 75 voters voted R, 25 voted D (75% R - 25% D)
- District B 75 voters voted D, 25 voted R (75% D - 25% R)
Now let's increase Dem turnout by 10 votes in each of those districts:
- District A 75 voters voted R, 35 voted D (68.2% R - 31.8% D)
- District B 25 voters voted D, 85 voted R (77.3% D - 22.7% R)
In District A, the Republican margin decreased from a 50% margin to a 36.4% margin - what looks like a swing of 13.6%, because of how percentages work mathematically (ceiling of 100%, floor of 0%).
Whereas in District B, the Democratic margin increased from a 50% margin to a 54.5% margin - what looks like a swing of 4.5%, because of how percentages work mathematically (ceiling of 100%, floor of 0%).
So if you are just looking at swing as you probably normally would (looking at % in a particular county or a particular district), it looks like the swing in District A is bigger/more impactful.
But in terms of the national popular vote (and the national popular vote %), the increase in turnout in both districts has the same effect. 1 additional vote is 1 additional vote.
Similar sort of math/simple examples apply for actual changes in who a voter is voting for (as opposed to turnout changes).
