Though of course it means little if Dems don't keep working and don't turn out to vote, and don't get their votes counted, here's an interesting
notes from Paul Krugman on his Friday column
"Will the Levee Break?" (sub.req.) and how he came to his conclusions:
. . .
My back of the envelope calculation ran as follows. I used the 2004 data to calculate the Republican share of the combined R and D vote for each district in 2004 (counting Independent Bernie Sanders as a Dem), then ranked the districts and graphed the result, which looks like this:
In 2004, every seat to the left of the point where the curve crosses 0.5 went to the Republicans. Now suppose that what happens in this election is that a certain percentage “x” of the voters shifts from R to D in every district – a crude model of the change in the political winds, but probably as good an approach as any. This will let the Dems climb the curve: every seat for which the Republican share in 2004 was less than 50% + x will shift parties. But as you can see, the curve is very steep where it crosses 0.5, so even a sizeable x doesn’t gain the Dems many seats. To reduce the R’s to 217 seats, x has to equal about 5 percent, which would correspond to an overall Democratic lead in the vote of about 7 percent.
But here’s the thing: if the Dems climb to that point, they’re past the steep “wall” that crosses 0.5 — the levee that protects the Republicans — and even a little bit more gives them big gains. If x equals 7.5 percent, corresponding to a 12 point generic lead — Franklin’s estimate — I have the Dems gaining 30 seats.
. . .