2018 Congressional Generic Ballot and House Polls Megathread - the original

[Torie]

"Yep, it's a disaster for the Dems. Considering that they've banked on winning suburban seats like Wagner's...this will be a rough November. I can see the GOP gaining 10 seats at this rate. Just wait for the N. Korea bounce."

Virginia, I am one who is hostile in general to the trolling rap (no secret to anyone I suspect), but the comment above is LOL trolling (nobody can be that profoundly ignorant who is a political junkie in good faith), and you have made a good case it is part of a consistent pattern, so I have more sympathy if you pursue this poster on that ground, then I generally would have. It is time that he toned it way down to a dull roar, or at least make his trolling more entertaining and creative. Hofoid, my best advice to you if you read this is to rein this in on your own ASAP.

[Torie]

The House is a tossup as to which party controls it after the next election. Beyond that finding that who ends up controlling the House is a coin flip (which does not surprise me - I suspect the odds that the Dems win it are lower than what most around here think), of more interest to me, is what folks think about their methodology.

Their approach involves polling cohorts of voters based on demographics, and then calculates their respective numbers in each of the CD's deemed potentially competitive (about 60 of them), and then using a weighted average appropriate for a given CD of the cohort poll results, comes up the the odds in each CD of one party or the other winning the seat, and then runs Monte Carlo simulations, to come up with the over/under estimated partisan split of the House overall.  I think I have paraphrased that right, more or less, but you can be the judge of that.

[Torie]

Well screw the poll and its fancy metrics. Let's keep the auction going. I hear 80% from the Massachusetts liberal. 80% going once. Ah, 60% from the California Commie. Thank you CC! Anyone here willing to go for 55%? How about 58%? Going once ...

[Torie]

Estimating these sort of probabilities is not a particularly exact science.

I don't think sort of modelling approach is very precise, given that you'd expect demographics to vote very differently between different districts (ok not VERY VERY differently, but enough around the margins that it becomes a very noisy predictor).

I'd probably rate Democrats' chances at taking the House as a little better than coin flip but not that much higher. Maybe 60% or so.

That bid was already taken.

I think the game here is to finesse away the problem with small subsamples, which is what you would have on a CD by CD basis, by using a national sample that is finely grained by cohorts. But as you say, sometimes after correcting for everything else, geography still matters. Heck, there is a lot of chat these days, that folks actually choose to live with folks who politically agree with them. So the Godless gay white man living in a large metro area in a close in elegantly aged leafy upper middle class suburb who is a lawyer with X income aged 67 living in Dallas might have different political views than the otherwise demographic clone of him living in San Francisco. Perhaps they have a trip wire when big geographic discrepancies arise, and then cut the grain finer to try to help close such inconvenient gaps.

[Torie]

Because of gerrymandering, I don't think they'll win the House, but I think it's almost certain they'll win the Senate.

That's quite a prediction.

Particularly since the Senate lines are gerrymandered as it were. . I wonder if WV was divided from Virginia in order to make Virginia a Dem state about 150 years later.

[Torie]

PA-01 (Monmouth):

Fitzpatrick 49% - Wallace 42% (RV)
Fitzpatrick 48% - Wallace 47% (LV)
Wallace 48 - Fitzpatrick 47% (LV - Dem Surge)

Source

Usually the Pub does better rather than worse with LV's versus RV's.  That suggests a malaise problem for the Pubs, or maybe an indicator that the SES profile of Pub voters on average as fallen in the Trump era, or both.