Respectability Politics Strikes Back: College Degree Holders and the 2016 House

Submitted by Charlie on Thu 27 Oct 2016 - 11:34
Irish Suck - Vote for Duke

"Sir, the polls show you're doing great with voters across the board, except women." - Phillips for President Campaign Manager
"Do they vote?" - Duke Phillips
"Yes. We do." - Alice Tompkins
"Really? Well, what about the Irish?" - Duke Phillips
"Them too." - Alice Tompkins
"Uh-oh, better change these posters." - Duke Phillips

There has been a slight downturn in the House generic ballot since last week, with the Pollster average down to 4.6% and the RCP average at just 3.8%. (Remember, the magic number is 7.0%, but the 2012 polls understated the eventual Democratic vote by a point or two, so anything north of 5.0% is possible, and north of 6.0% is outright good.) However, there's a wrinkle that needs a lot more attention than its getting: Trump's collapse with college educated white voters.

Non-college whites have been the belle of this year's ball, with seemingly every reporter, pundit, and blogger writing about them. Minority voters have been under-covered (as usual), but there are only so many ways to write "brown people don't like Trump for lots and lots of very good reasons". But the biggest historical anomaly in this year's polls is the abandonment of the Republicans by white voters with college degrees.

White degree holders have been reliable Republican votes since time immemorial (or at least since modern polling got figured out). These are small business owners, dentists, engineers, office managers, etc., work-a-day white collar folks who want their taxes low, don't care much about social issues, and fear terrorism and crime with abandon. They've been trending away from the GOP for years as the Republicans increasingly became the "stupid party", but 2016 has seen the bottom fall out. 

In 2012, Romney won college educated whites by about 14 points (+21 with college educated men, and +7 with college educated women). This year, college educated women are heavily backing Clinton, and college educated men are frequently drawing even, which means this once reliable Republican voting bloc has crumbled. The ABC tracking poll that debuted Sunday is fairly typical of ones I've seen this year: Clinton is +32 with college women, and dead even with college men (0), for a total of +16 overall.

The big unknown for the House is how much of that flaming Trump wreckage is going to translate down to individual district races. In the absence of publicly available House polling, we've got to make some guesses, but the best place to start is the Census bureau's demographic information on House districts. Pulling from that, I've generated a spreadsheet with all the Republican held House districts that are rated as competitive by Kos, Cook, Rothernberg & Gonzales, and Larry Sabato, plus a few that were closer than they maybe should have been in 2012 or 2014. Sorting by highest percentage of people with a college degree, you get this:

Stay in school, kids, the you won't vote for fascistic loons.

If we're going to get a majority (unlikely but not impossible) or make serious inroads and get their majority to <10 seats (looking better all the time), some of these listed as Lean or Likely Republican should be ripe for the plucking. The national average for college degrees is 33%, so anything north of that is bad for Trump and (hopefully) bad for the Republican on the ballot as well. The big target at the top of that list is VA-10, which is a very Republican district that is nevertheless seriously in the mix this year because it's full of people with four year degrees.

But it's the races listed as Leaning Republican, Likely Republican, and "N/A" (which means they're considered Safe) that we need to keep an eye on. Trump is likely to do worst in these heavily college educated districts than elsewhere, but these races are still friendly to Trump for other reasons (primarily that they're wealthier and whiter). However, IF (1) there really is a wave going against the Republicans across the board, and IF (2) Republican turnout is down generally, and IF (3) that wave crests among college educated white people (which polling indicates is likely), then these are the districts that are most likely to tip the House (or at least get us close enough to pass better legislation than has been possible since the 2010 disaster).

So if districts like VA-10 and NJ-05 go Blue early on Election Night, that's a very good sign. If districts like MI-11 and PA-06 don't get called quickly, that's also a very good sign.

Beyond this year, these numbers are also going to be the first piece of data we get about whether or not becoming the party of Trump and stupidity has hurt the Republicans once Trump is off the stage. If it's a Trump blip, then in 2018 and 2020, college educated white people will return to their old ways. If Trump isn't a blip, but rather the hair plugs that broke the camel's back, then these patterns will remain, and suburban Republicans will find their prospects a lot dimmer than they used to be for years to come.

Given the Trump Meltdown, Is the House Winnable? (Probably Not But Maybe Kinda Yes)

Submitted by Charlie on Mon 17 Oct 2016 - 11:41

"I'm worse than Hitler?" - Jay Sherman
"Not worse, just less warm and cuddly. Jay, test groups like this determine what you watch on TV, what kind of car you drive, even who runs this country." - Duke Phillips

Back in July, I took a detailed look at the House map and concluded that the Democrats had only the longest of long shots to retake the House. Three months and one Donald Trump poll implosion later, and the odds have improved enough that it actually makes sense to take another look. So, how hopeful should we be for a Blue House majority?

First, there are three (3) big unknowns to overcome. 1) House polling is expensive, and doesn't lend itself well to clickbait articles, so there is very little publicly available polling on individual House races. The main tool we have is the "generic ballot" question, which is usually tacked onto other polls and used as a stand-in for the House race overall. This means we simply don't have the data to make a prediction on a district-by-district basis, so we have to use the generic number as a stand-in.

That brings us to problem number 2), there isn't a 1:1 correlation between the overall House vote and the number of seats each party wins. Because each district is its own self contained contest, an increase of 1% in the Democratic vote nationwide might translate into anywhere from zero extra seats to five or more depending on where those votes come from. As a further wrinkle, Republican controlled seats tend to come from much whiter districts (duh), so the shift in the white vote will affect Democratic House outcomes a lot more than a surge in Latino voting or 2008/2012 turnout levels among black voters.

Muddling matters even further is number 3): this will be the first Presidential election run after the grotesque Shelby County decision. Since 2013, a slew of Republican controlled states have enacted all manner of voter suppression laws, and many of them are actively fighting and ignoring court orders to drop them as we speak. We can't know how many, but we can be damn sure that thousands, perhaps even hundreds of thousands, of Democratic votes will be "missing" because of reductions in early voting, restrictive ID laws, and similar fiendery.

TL;DR for the above is that we're trying to use a general number to account for hundreds of distinct contests, each of which has its own complications based on local politics, voting restrictions, and gerrymandering. So the best we can do here is make some educated guesses.

The baseline we're going to start with is the 2012 election, which saw the good guys win a 1.2% overall victory in terms of House ballots (a margin of 1.4 million votes). Despite that, the resulting House was a Republican blowout of 234-201 in terms of seats. That's the hole we have to climb out of. Our own Stephen Wolf puts the likely margin needed at 7-8 points which accords with what Harry Enten (now of 538) thinks.

The generic ballot polling in 2012, especially once it got into October (when polling accuracy tends to get much better) was pretty close to the actual outcome, with a slight Republican lean. The polls from October forward ranged from +3 Republican to +2 Democratic, with Real Clear Politics placing their final weighted average at +0.2 Republican, which was 1.4% off from the actual outcome.

The current RCP average is +5.0 Democratic, and the polls that produced it range from +3 to +10. That's probably not going to be enough, but if the Democrats out-perform their polls by the same 1.4% margin they did in 2012, that would put the margin at +6.4 Democratic, which is still short of 7, but close enough that immeasurable factors could flip the chamber. Plus there was a Politico/Morning Consult poll this morning that isn't in the RCP average yet. It was +7 Democratic.

Given all the fudge factors above, I'm not going to assign a probability to whether or not the Democrats can take the House. Any number I give would be pulled straight out of my ass. But what we can say is that if the polls hold roughly where they are, the Democrats have a puncher's chance, which is a hell of a lot better than things looked over the summer.