State Analysis: Texas, Nevada, and incorporating turnout into elasticity

2020 turnout is expected to be sky-high, and elasticity as constructed doesn't truly account for county growth and the potential for turnout altering the nature of the election. However, you can extend elasticity to account for turnout variation as well, and this can give you a better idea of "target counties" in which heavy investment is needed, whether for flipping swing voters or registering new ones.

With the help of Sawyer Jack Robertson (@Sawyer_Jack_), I extended elasticity to add in turnout by extending the (Democratic, Republican) plane to (Democratic, Republican, Turnout), where Turnout is the percentage of the county's population (according to 2010 census data) that cast a vote. We then take the pairwise euclidean distance between all elections in the set, as discussed in my initial article on elasticity here.

Under this lens, Texas becomes a genuine tossup.

Texas

Texas' Elasticity without turnout variation

Above, we see Texas' elasticity with only swing voters. The state is "difficult, but winnable" -- there are enough metro areas that are elastic to see Biden flipping enough suburban voters to take the state, but it's a tough task.

Texas' Elasticity with turnout variation

However, when factoring in turnout variation, we see that, along with general increases in elasticity across the board, the population boom in the Austin area has led to droves of new voters, which is excellent news for Democrats and expands their victory horizon. In particular, Travis (E 68.7), Williamson (E 69.7), and Hays (E 78.0) are key areas that Democrats will look to grow their vote totals in.

Another key set of counties to keep an eye on are the urban areas of Collin (E 65.1), Tarrant (E 41.0), Dallas (E 44.9), and Denton (E 62.1) -- Biden will need big wins here to carry the state in 2020, but the rapid growth in Collin and Denton should serve him well. Evidence of this growth can already be seen in comparing the 2016 Presidential election to the 2018 Senate race; while Clinton lost Collin by 17 and Denton by 20, O'Rourke only lost these counties by 6 and 8 points, respectively, in 2018.

Don't get it twisted -- swing voters are still a huge key to winning Texas. The Rio Grande Valley, in particular, has many conservative Democrats and is extremely elastic, and Biden needs to address his purported weaknesses with Hispanic voters and hold down that area to ensure it doesn't offset his continuing growth in the suburbs.

But when turnout enters the equation (and, as we are seeing, it is rapidly increasing in Texas in 2020), the state becomes way more gettable and goes from "tough, but winnable" to "tossup". It's well and truly in play, because the Democratic areas are gaining a lot of new voters. The surge in turnout we're seeing reflects the elasticity of the areas; Collin and Denton, in particular, have already cast over 35% of their votes with 12 days to go until the election. Stunningly, Harris county (E 43.5) has cast over 50% of its 2016 vote with two weeks to go until the election, and the gains in votes are reflected among young voters and women, two groups that break strongly Democratic. If these are indicators of the results we will get on November 3, it's hard to imagine the race as anything but a tossup, and I would go so far as to call it one in which Democrats are the ever-so-slight favorites.


Rating: Tossup/Tilt Democratic.



Nevada

Nevada's Elasticity with turnout variation


There's some discourse about Nevada being in danger of flipping, but if this is the claim, you'll have to use something stronger than "well, it was close in 2016!". It's not the vote margin -- it's the fact that this state has very little room to move. Over 90% of the state lives in inelastic counties, and that makes the math extremely tough for Republicans going forward.

This map accounts for the growth of Clark County (E 25.0), where Vegas and its suburbs lie -- 73% of Nevada's population calls it home, and it is the state's fastest-growing county. Hillary Clinton and Catherine Cortez Masto won it by 11 points, and Jacky Rosen won it by 14. However, it's quite inelastic, and that isn't great news for Republicans; it means the electorate there may be less prone to massive swings in wave elections without a fundamental coalition shift, which is quite good for the party in power.

Meanwhile, Washoe (E 23.8), the next largest county, with about 12% of the total population, has a small Democratic lean. Washoe and Clark combine for about 85% of the total population, so that makes the math extremely favorable for Democrats despite the rest of the state being fairly Republican -- it means their electorate and vote margins are fairly locked in and less prone to the swings you may see in midterms and wave years. This also bodes well for their chances of holding the state in the future.

Inelasticity means that a five point lead in Nevada means much more than a 5 point lead in Wisconsin. Do not look only at the polling averages -- look at them in context of the state's elasticity. Nobody is claiming Joe Biden should take the state for granted, but Trump going on the offensive here recently wasn't part of a five-dimensional chess strategy, because Nevada is not a state you would expect him to make gains in come 2020. It's only because he's flailing in the water and grasping at anything he can to keep his campaign afloat. He's trying to keep the playing field as broad as it was in 2016, and he sees Nevada, a state where he came close in 2016, as one of the few ways to do so. Nevada is a likely Democratic state come November, but it's a lot closer to "safe D" than it is to "lean D".

Rating: Likely Democratic


A special thanks to Sawyer Jack Robertson (@Sawyer_Jack_) and Steph Hernandez (@StephieTheLefty) for their invaluable assistance with this -- Sawyer for helping to add turnout to the metric, and Steph for their analysis with Texas.

Comments

  1. In a 3D space like Texas, you might want to consider using Mahalanobis distance instead of Euclidean distance to measure elasticity as the correlation between variables is removed in the former.

    ReplyDelete
    Replies
    1. Interesting idea, thanks! I'll look into it; never really considered this before.

      Delete

Post a Comment

Popular Posts