Obviously the presidential polling continues to report favorable outcomes for the Dems. And I've enjoyed reading the different analysis (albeit a bit leftist) on 538. However, I pulled one of the most recent polls that actually gave some good participant information. Poll used is the Emerson College poll. Again, mostly because it does such a thorough breakdown of its respondants. Many of the other polls don't.
But in each of the 5 states analyzed (FL, NV, OH, NC, & WI?) am I missing something or did they over-sample registered democrats? Or do these percentages align with past voting patterns or through some math?
Results:
FL: Hillary +1
OH: Even
WI: Hillary +6
NC: Hillary +3
NV: Hillary +2
But check out these splits:
Florida actually is reasonable: 170 Registered. Dems;165 Reg. Repubs; 165 Independents
However, they have a 281 to 219 (56% to 44% split female vs. male). This seems a bit high to me, and would lean a poll towards Hillary.
OH: 296 Dems vs. 252 Repubs vs. 252 Independents? Does a 37/31.5/31.5 split align to Ohio's registrations or voting patterns, or is this not clear oversampling? They did at least balance the gender portion out to be what I would expect 51/49 female.
WI: This actually looks reasonable: 36/34/30 split on D/R/I registration. 51/49 F/M split.
NV: 38/30/32 D/R/I registration? If this spread is even say 34/34/32, wouldn't you most likely see Trump +2? Also had a 53.3/46.7 F/M split, which seems to be on the high side. If nothing else, with this demo split, I'd fully expect Hillary to be up more than 2%.
NC: 39/33/28 D/R/I split and a 55/45 F/M split. This feels like a very pro-Clinton sampling.
I guess my question is what am I missing? It feels like pretty significant pro-Clinton sampling, but maybe there's justification for consistently sampling more registered Dems is states like NV, OH, and NC?
Justin
Edit: Meant to include link to the data: http://projects.fivethirtyeight.com/polls/20161029_1.pdf
But in each of the 5 states analyzed (FL, NV, OH, NC, & WI?) am I missing something or did they over-sample registered democrats? Or do these percentages align with past voting patterns or through some math?
Results:
FL: Hillary +1
OH: Even
WI: Hillary +6
NC: Hillary +3
NV: Hillary +2
But check out these splits:
Florida actually is reasonable: 170 Registered. Dems;165 Reg. Repubs; 165 Independents
However, they have a 281 to 219 (56% to 44% split female vs. male). This seems a bit high to me, and would lean a poll towards Hillary.
OH: 296 Dems vs. 252 Repubs vs. 252 Independents? Does a 37/31.5/31.5 split align to Ohio's registrations or voting patterns, or is this not clear oversampling? They did at least balance the gender portion out to be what I would expect 51/49 female.
WI: This actually looks reasonable: 36/34/30 split on D/R/I registration. 51/49 F/M split.
NV: 38/30/32 D/R/I registration? If this spread is even say 34/34/32, wouldn't you most likely see Trump +2? Also had a 53.3/46.7 F/M split, which seems to be on the high side. If nothing else, with this demo split, I'd fully expect Hillary to be up more than 2%.
NC: 39/33/28 D/R/I split and a 55/45 F/M split. This feels like a very pro-Clinton sampling.
I guess my question is what am I missing? It feels like pretty significant pro-Clinton sampling, but maybe there's justification for consistently sampling more registered Dems is states like NV, OH, and NC?
Justin
Edit: Meant to include link to the data: http://projects.fivethirtyeight.com/polls/20161029_1.pdf
