Qunnipiac: Trump +5 in TX, Biden +1 in OH

[wbrocks67]

Q has the same issue with Latinos in TX as FL. Biden +8?? Most likely Trump is better with whites than in this poll but a lot worse with Latinos.

Yeah, was just gonna say. For some reason, Q-pac has a huge problem contacting Latinos. But I think this sample is just super R-leaning in general. Biden has the support of 95% of democrats and an +8 lead among Indies (Beto only won them by 3) and he's still losing by 5. The "likely voter" sample seems to benefit Rs a lot here. Plus, they have Trump at 19% with blacks and 43% with Hispanics which both are really out of step.

I'm not really convinced Ohio is this close just b/c we've been burned here before, but Biden is leading +1 here even while losing 18-34 year olds which is obviously not gonna happen. So who knows.

[wbrocks67]

Q has the same issue with Latinos in TX as FL. Biden +8?? Most likely Trump is better with whites than in this poll but a lot worse with Latinos.

To add to this, just like in 2018 when they severely underestimated Beto in all of their polls:

Cruz +5, Cruz +9, Cruz +9, Cruz +6, Cruz +11

[wbrocks67]

Although if you think about it: it may not be that surprising. Hasn't Biden been up with ads in Ohio for a while now? While Trump is too broke to even spend $$$ there. If this is accurate, that could be a major reason why.

[wbrocks67]

I don't believe either of these results. Also, Quinnipiac needs to get better at polling Black and Latino people.

It appears that it's limited to the sun-belt stats. They have Ohio at 91% Biden with blacks and 6% Trump which is probably very close to the likely result. For some reason though, in places like TX, FL, AZ, they have a really hard time with it. Or a bad LV model.

[wbrocks67]

RCP 2018 average: Cruz +6.8 > off by 4.2% towards D
583 2018 final forecast: Cruz +4.9-5.3 > off by 2.3-2.7% towards D

538's poll average right now is Trump +2.0 and their forecast is Trump +4.0. If we see the same 2-4% problem again, Biden wins.

I wish Q-pac released their RV # too. I have a feeling the LV model is the problem here.

[wbrocks67]

RCP 2018 average: Cruz +6.8 > off by 4.2% towards D
583 2018 final forecast: Cruz +4.9-5.3 > off by 2.3-2.7% towards D

538's poll average right now is Trump +2.0 and their forecast is Trump +4.0. If we see the same 2-4% problem again, Biden wins.

I wish Q-pac released their RV # too. I have a feeling the LV model is the problem here.

By this definition, WI is tilt Trump as he is  down 6.6 points in the RCP average vs 6.5 last time and if you apply same polling error than Trump wins WI.

This, 2016/2018 polling errors ≠ 2020 polling errors

My point is that in general, pollsters tend to underestimate Ds strength in states that heavy latino / minority populations like CA, AZ, TX, etc. Quinnipiac however specifically has this issue as well. We all know Trump is not getting 20% of the black vote and 43% of the latino vote in TX.

[wbrocks67]

The Texas sample is 34 R/28 D/38 I.

This is likely too Republican-friendly, as 2018 turnout looked like 38 R/34 D/27 I.

One could "unskew" by plugging the crosstab data into a different turnout model, but that would be mostly futile since the poll's other weights are probably impacting the crosstabs. My best guess is that a R+4 sample with Biden leading independents by high single digits would produce a tie or 1-point lead either way.

The gold standard has spoken.

This lines up with what Nate Cohn said earlier, about if you apply the strictest "most likely" LV voter screen to his poll that most pollsters do, you would also get about a tie in the TX race.

[wbrocks67]

lmao

Biden winning Hispanics 51/43? This is no way the real result in November is that close.

Not to mention, if Biden is winning Indies by 8 (versus Beto +3 and Trump +14), he's def not down by 5 statewide.

[wbrocks67]

RCP 2018 average: Cruz +6.8 > off by 4.2% towards D
583 2018 final forecast: Cruz +4.9-5.3 > off by 2.3-2.7% towards D

538's poll average right now is Trump +2.0 and their forecast is Trump +4.0. If we see the same 2-4% problem again, Biden wins.

I wish Q-pac released their RV # too. I have a feeling the LV model is the problem here.

By this definition, WI is tilt Trump as he is  down 6.6 points in the RCP average vs 6.5 last time and if you apply same polling error than Trump wins WI.

This, 2016/2018 polling errors ≠ 2020 polling errors

This is starting to look different from 2016/18 in a meaningful way.  Latinos are swinging toward Trump while everyone else swings away.  IDK about polling errors though.

Eh, I think it's just pollsters across the board struggling to poll Latinos, which has been in issue for years now. There is absolutely a possibility of Biden doing a little worse than Clinton among them, but given that they went about the same amount in 2016 and 2018 for Dems, it wouldn't make much sense for a huge change to occur, plus the fact that you have some polls with great #s for Biden among them and others with terrible.

[wbrocks67]

https://twitter.com/NateSilver538/status/1309198520394625024

Nate's really trying to play pundit here, yet ignoring that other than ABC/Wapo poll for AZ (that also had Biden +2 among RVs!), the polls have still been largely good for him there, in a state *Trump won* in 2016. Suddenly there is such a high bar where even if Biden is outperforming Trump by 4/5 in a state, it's still not good enough or "mediocre"

We literally just had the +9 NYT poll late last week.

[wbrocks67]

Well, at least their TX poll was accurate lol