The state of the polls

Hillary Clinton

I’m just trying to get a handle on the numbers, with the idea of establishing some kind of guide for what to expect in the Presidential race in Texas. Bear with me.

The RCP average for the two-way Trump/Clinton race is 44.0 for Trump and 38.3 for Clinton. The FiveThirtyEight polling averages, which includes some other sources, come in at Trump 45.6, Clinton 37.6. However, once you apply the 538 secret sauce, you wind up with projected totals of 49.7% for Trump and 43.2% for Clinton.

RCP does not do this kind of modeling/forecasting – it’s a straight up polling average. As such, it can underestimate final totals, since it doesn’t try to guess what undecided voters may do. The 2012 RCP average for Texas had President Obama at 39.0 and Mitt Romney at 55.7; they finished at 41.4 and 57.2, respectively. Similarly, in 2008, Obama was averaging 40.5 and John McCain was at 53.5; the final numbers were 43.7 and 55.5. In other words, RCP underestimated Obama by three points in 2008 and by 2.5 points in 2012.

(I couldn’t find 538’s data for Texas in past years, so we’ll just skip that part of the analysis.)

There are so many variables in play here that I’ve been very reluctant to even begin to guess at what the final numbers might look like. Here are some of the things that factor in:

1. Overall turnout – Voter registration is at an all-time high, but that correlates weakly at best to turnout. However, the overall voting age population is way up, and even in a modest turnout-to-VAP scenario like we had in 2012, we’re easily looking at a half million or more extra voters than we’ve ever had, and that number could be quite a bit higher without setting a record for turnout as a share of the adult population. Nine million votes is not out of the question. I have to believe that beyond a certain point, extra voters will break Democratic. Where that point is, how blue they are, and how likely that is to happen, I have no idea.

2. Undecided voters – In 2008, the Obama/McCain share of the vote in the averages was 94.0%; in 2012, the Obama/Romney share was 94.7%. This year, it’s 82.3% for Trump and Clinton. Even adding in Johnson and Stein only gets you to 91.6%. That’s a lot more undecided voters. Do they show up? Which way do they lean? There’s a lot of room for candidates to gain ground here.

3. The third-party candidates – Just as a reminder, Gary Johnson and Jill Stein combined for 1.42% of the vote in Texas in 2012. Their RCP combined average is 9.3% right now. Poll numbers for third-party candidates are almost always overstated, often by quite a bit, but we don’t have any useful data for comparison from 2012. I’m sure there are some Republicans who will vote for Johnson over Trump, but nearly the entire state GOP establishment is in Trump’s corner, so it’s not like there’s an organized #NeverTrump movement. As with the undecided voters, there’s a lot of room for the Trump and Clinton numbers to change here if as has been the norm historically the L and G numbers are exaggerated. But if there was ever a year where maybe they’re not, you’d think this would be it.

4. The other polls – There are national polls showing Hillary Clinton with a double-digit lead. That’s a landslide by any measure, and if it’s what we get, it’s entirely possible that the polls we have for Texas are underestimating her by a considerable amount, as state polling tends to lag the national trends. The fact that the one most recent poll we have is also the closest one we’ve seen since that weird Washington Post poll suggests that possibility as well. We also know that there’s a lot of polling data that is not made public but from which we can make inferences based on the actions taken by the campaigns and other actors who have that data. Here, we have multiple suggestions of Republicans being worried about their turnout in Texas, plus Hillary Clinton actually running a week’s worth of ads in Texas, online and on TV. Draw your own conclusions about that.

5. Latino voters – This is baked into some of the other factors, but I keep being struck by the differences between what national polls say about Latino support for Donald Trump – in short, he may be lucky to get 20% of the Latino vote nationally, well below what Mitt Romney got – and what the state polls have said. The latter have generally had his support in the 30s, with Clinton in the 50s or low 60s. This may be a function of small sample sizes combined with excessive weighting to compensate, or it may simply indicate that Texas Latinos are different than Latinos elsewhere. Bear in mind that we have some data to indicate that lower-propensity Latino voters tend to be more Democratic than high-propensity Latino voters, which is a fancy way of saying that higher Latino turnout correlates with better Democratic performance among Latinos.

6. Crossover voters – Mark Bluenthal wrote yesterday that the key to Hillary Clinton’s increased national lead is that she has consolidated the Democratic vote better than Donald Trump has done with the Republican vote. Another way to put that is there are more Republicans who are voting for other candidates, including Clinton, than there are Democrats who are voting for other candidates. We see that in Texas as well, specifically in that UH poll, which showed ten percent of Rs voting for Clinton or Johnson, but only five percent of Ds voting for other candidates. Hillary Clinton’s better performance in Texas is two parts turnout – there are more Democrats and fewer Republicans voting than usual – and one part crossover voting. If that latter group is bigger than we think, that will affect the outcome.

In the end, I’m less interested in the margin between Trump and Clinton – given what we do know so far, barring anything unexpected that margin is going to be smaller than the McCain-Obama margin – as I am in the absolute totals. How many people actually vote for Hillary Clinton? The high-water mark is 3,528,633, set by Obama in 2008. Just on the increase in population alone, she could top that while receiving a lower percentage of the vote (for example, 3.6 million votes for Clinton out of 8.4 million total = 42.9%; Obama got 43.7%), but I would consider that a huge disappointment. Can she get to 3.8 million, or (be still my heart) 4 million? Can she reach 44 or even 45 percent, a level not reached since Jimmy Carter in 1976? I hope to have some small amount of clarity on this before voting concludes, but I doubt I’ll get much.

I think that about covers it. What it all means, I still don’t know. But when it’s all over and we’re doing the autopsy, these are the things I’ll want to look back on.

Related Posts:

This entry was posted in The making of the President and tagged , , , , , , , , , , , , , , , . Bookmark the permalink.

One Response to The state of the polls

  1. Paul Kubosh says:

    Excellent post. This is why I stop by here nearly every morning. Good job.

Comments are closed.