CNN/ORC: CO: Trump +1 PA: Clinton +1

[TheDeadFlagBlues]

In all likelihood, Clinton doesn't have anything scheduled in Colorado because her analytics team doesn't think it's all that competitive. I sound like a broken record right now but these results aren't compelling to me; ~43% of Colorado Whites have a Bachelor's degree or higher, there's been a massive influx of kush-smoking millennials to Denver, there's a very substantial Mexican-American community in Colorado etc. Pollsters also have a perennial problem estimating Colorado's voting behavior. They're always off in the Republican direction. There's every reason to believe that this problem has only gotten worse recently.

I can buy PA to a degree but there's no way that CO is going to be closer than PA unless Clinton wins in a blowout imo.

[TheDeadFlagBlues]

In all likelihood, Clinton doesn't have anything scheduled in Colorado because her analytics team doesn't think it's all that competitive. I sound like a broken record right now but these results aren't compelling to me; ~43% of Colorado Whites have a Bachelor's degree or higher, there's been a massive influx of kush-smoking millennials to Denver, there's a very substantial Mexican-American community in Colorado etc. Pollsters also have a perennial problem estimating Colorado's voting behavior. They're always off in the Republican direction. There's every reason to believe that this problem has only gotten worse recently.

This ignores the fact that the 2014 senate result was IDENTICAL to the RCP average.

I don't think you understand how this works son. Polling elsewhere was systematically biased in the Democratic direction. In this case, the fact that Colorado's 2014 senate result was identity to the RCP average is actually an indication that Colorado pollsters have a hard time obtaining representative samples. As a counter-example, Hickenlooper outperformed his polling figures by a bit over 2 percentage points.

IIRC, most people on Atlas commented on this in 2014. In retrospect, it's astonishing that Colorado was so close because the bottom fell out of the Democratic Party fell out everywhere else.

[TheDeadFlagBlues]

Colorado 2012: O + 5.36
Pennsylvania 2012: O + 5.38

Colorado % of Non-Hispanic Whites with Bachelor's or Higher: ~43%
Pennsylvania % of Non-Hispanic Whites with Bachelor's or Higher: ~29%

To re-state my case: there is every reason to be skeptical about this set of polls because the idea that Colorado is going to vote two points to the right of Pennsylvania this year is inane. I'd argue that it's not unreasonable to expect Colorado to be close in the event that this election is a tossup but it's certainly unreasonable to expect it to be more favorable to Trump than Pennsylvania.

Look, I get it, this election is very stressful and the polling volatility is panic-inducing but these polls don't pass any kind of smell test. If I had the ability to make a candidate in a lab who was bound to lose Colorado, he'd be very similar to Donald Trump, who has managed to alienate every prominent constituency in the state. This doesn't even take demographic changes into account or changes in voter registration statistics; I'd expect Trump to lose Colorado by a large margin under conditions of a 2012 voter universe. 4 years have passed and Colorado's likely universe is quite different now than it was then. It's younger, it's a bit more Hispanic and it's more liberal then than it is now. There are tremendous age disparities in Colorado that are quite similar to shocking age disparities seen in, say, North Carolina or Arizona.

I'm very sleep deprived so I apologize if this post isn't all that coherent but I can't see how I'm wrong about Colorado. If anyone wants to pose a serious critique to my argument, I wouldn't mind engaging in a back in forth. It's entirely possible that my blinders are on right now...

[TheDeadFlagBlues]

It's getting to the point we can summarize many poll reactions this way:

"My preconceived notions of how people who disagree with me think must be superior to this undesirable polling result."

At this point, I see a lot of people basically saying, "PVI is permanent and perfect."

There is an argument that Trump is the closest thing to a map-scrambling candidate we've seen in our lifetimes.  Add that to the fact these are the most two disliked candidates in recent memory, and that's a recipe for some significant PVI-busting oddities this year.

And we have quite a few polls that are confirming that.  It's not impossible.

This is a pretty straightforward question: do you think it's plausible to argue that CO's PVI is going to shift to the right of PA's PVI? If so, why do you believe that this is plausible?

[TheDeadFlagBlues]

It's getting to the point we can summarize many poll reactions this way:

"My preconceived notions of how people who disagree with me think must be superior to this undesirable polling result."

At this point, I see a lot of people basically saying, "PVI is permanent and perfect."

There is an argument that Trump is the closest thing to a map-scrambling candidate we've seen in our lifetimes.  Add that to the fact these are the most two disliked candidates in recent memory, and that's a recipe for some significant PVI-busting oddities this year.

And we have quite a few polls that are confirming that.  It's not impossible.

This is a pretty straightforward question: do you think it's plausible to argue that CO's PVI is going to shift to the right of PA's PVI? If so, why do you believe that this is plausible?

CO's PVI was to the right of PA's in 2010. They are currently both D+1 states.  Why wouldn't one state shift more to the right of the other?  PVI doesn't necessarily move in lock step.

Oh, I buy that it's plausible for one state's PVI to the shift to the right of another state's. That's the nature of PVI, no? My argument is that it isn't plausible for CO's PVI to shift to the right of PA's PVI. That's entirely implausible based on every piece of data we've seen thus far, where there's been a tremendous shift of college-educated whites to Clinton and a tremendous shift of non-college whites to Trump. Every indication is that this is the defining characteristic of this election. As such, it strikes me as preposterous to think that Colorado, a one of the most educated states in the country, could vote to the right of PA. If we were to construct a model based on crosstabs of national polls that predicted the margin of victory in each state based on demographics, if such Colorado was to vote to the right of Pennsylvania, the results of both states would have insane residual values. They'd be outliers!

Drilling down further, let's suppose such a PVI shift (CO D PVI < PA D PVI) were to happen, what could explain it?

Note: I don't think that piece of evidence you've supplied is useful. Yes, Colorado's PVI was to the right of Pennsylvania's in 2010. Colorado was also a different state in 2010 than it is in 2016. The nature of partisan coalitions has shifted. If anything, the voting histories of both states are more misleading than they are useful.

[TheDeadFlagBlues]

Drilling down further, let's suppose such a PVI shift (CO D PVI < PA D PVI) were to happen, what could explain it?

One way is the college whites that will swing away from Trump to Clinton mostly goes to Johnson in CO but goes mostly Clinton in PA overcoming the fact that there are more college whites in CO than PA to swing.   So what seems like a plus for Clinton in CO relative to PA ends up being neutral or even negative. 

That's somewhat similar to my explanation, which is based on the fact that the age distribution in CO skews relatively young. If millennial turnout both plummets and, of those who turnout, more vote for Johnson/Stein than usual, that could shift CO to the right of PA. How plausible is this scenario? I'm not sure but it strikes me as being very implausible.

[TheDeadFlagBlues]

Cook's 2016 PVI projections for these states is: PA D+2, CO D+1.  (For comparison, 2012 he had it: PA D+2, CO Even.)

One point difference.  I'll repeat that...one point.

To say that it's impossible for them to flip is assigning a precision to demographic-based voting tendencies that doesn't exist in the real world.  Humans are complex.  Elections have all sorts of cross-currents.  PVI is an estimate of a trend.  It's a 2nd derivative.

CA to the right of UT is impossible.  CO 20 points to the right of PA is impossible.

CO 1-3 points to the right of PA is certainly possible.

"WV 1-3 points to the left of North Carolina is certainly possible." - some guy in 2008

You can't rely on historical benchmarks this year in the way that you could in 2012, which is why I discount Cook's predicted PVIs. Why bother utilizing them if we have ample national polling data that can be used to build models that can predict forecasts/trends?

I'd note that I haven't done this (yet) but it doesn't take a statistician to notice that, huh, if there's a massive swing to Hillary among college-educated whites and a massive swing to Trump among less-educated whites that Colorado would vote to the left of PA.