NATIONAL SENATE/HOUSE RESULTS THREAD (LATE RESULTS/POSTMORTEM)

[Alcon]

Sestak is running about 3% behind Obama in Philadelphia, and not doing that well in Berks County (part of Toomey's district) either...but those seem to be the exceptions.  Sestak seems to be performing rather well.  Bizarre?

[Alcon]

I'm looking at the Washington returns and it looks like Murray will win by about 4% and may surpass 100k over Rossi.

At least.  She has a shot at 5%.

[Alcon]

Harry Reid ended up winning Searchlight by just 3 votes.  

Only 3 votes ? That's certainly one of the poorest "home turf effects" ever.

Searchlight is pretty darn Republican.  I don't think Obama won it

[Alcon]

Not to mention that exit polls aren't truly random samples to begin with, so it's not just a margin-of-error thing.

[Alcon]

That's not a terribly implausible scenario, as there are about 5 black people in Vermont.

But don't you know that subsamples in polls are ALWAYS ALWAYS ALWAYS RIGHT NO MATTER HOW SMALL!!!!11!!!! [/Dgov]

Do I honestly have to make that clear before every post?  I kind of assumed that the people posting here would have enough common sense to know that when I'm using exit polls, I'M USING EXIT POLLS.  Your mileage may vary.

That doesn't really matter.  Exit polls filter out likely voters who don't vote, but are more prone to non-representative sampling, and either way subsamples in exit polls are basically just as bad as with polls.

[Alcon]

^^

cinyc -- I understand that; my point was assuming that all things are equal.  Considering how small that subsamples like that are even on larger-sample exit polls, I assumed he believed that exit polls were inherently prone to less statistical noise.  Either way, even the given exit poll sample is not large/random enough to so strongly draw that conclusion, IMO.