All eyes are on the sunbelt 📊 May 3, 2020

Early polls show Texas, Arizona and Georgia within the margin of error for November

Welcome! I’m G. Elliott Morris, a data journalist at The Economist and blogger of polls, elections, and political science. Happy Sunday! This is my weekly email where I write about politics using data and share links to what I’ve been reading and writing. 

Thoughts? Drop me a line (or just respond to this email). Like what you’re reading? Tap the ❤️ below the title and share with your friends! If you want more content, I publish subscriber-only posts 1-2x a week.

Share


Dear reader,

Good evening friends. I sent last week’s newsletter out on a Monday, bucking a long trend of Sunday newsletters. Though I had planned to go back to the normal weekend schedule this week, time got away from me gain yesterday. At least there’s an upside; last week, many more of you clicked on the newsletter than usual. We could switch the newsletter to Monday permanently, I guess, if I’m trying to maximize readership. We’ll see. Chime in via email if you’d like. In the meantime, here are some thoughts on data and electoral politics.


All eyes are on the sunbelt

Early polls show Texas, Arizona and Georgia within the margin of error for November

Aside from a rough sense of which states will be competitive, it’s hard to know far in advance of a presidential election just how competitive certain states will be. And the stakes for such calculations can be very high. The difference between a 20 and 40% probability of victory is huge when, for example, advocacy groups are deciding how to spend hundreds of millions of dollars.

Polls can be useful here. They help us figure out how things have changed since the last election. In our era of partisan hyper-polarization, even surveys fielded this far out from Election Day can be rather predictive of the actual results, and they certainly can help us gain an understanding of the distribution of ultimate outcomes.

I performed an exercise to this end over the weekend and wanted to share some findings with you all. The toplines are that Texas, Arizona and Georgia—long considered reach states for the democrats—have shifted to the left over the past 4 years and are just as competitive, if not more competitive, than the Midwestern states that decided last year’s contest. Of course, this is highly dependent on the national environment. Joe Biden is up by 6-7 percentage points nationally today. If his margin shrinks to 3-4 points by November 3rd, the lay of the land will shift too. But I digress…

I figured this out by calculating a relatively simple average of the last 2 month’s of polls at both the national and state level. I just took an average of each state’s polls, with each individual survey weighted by the number of people it sampled.

There’s one obvious hiccup to this exercise: not every state has yet been polled. So I trained a simple linear model to predict current polling averages in each state given its 2016 Clinton margin, population density and share of voters that are working class whites. This let me predict polling averages for each state. The final “average” in each state is a blend of the regression’s prediction, which gets assigned a weight equal to that for a poll with an average sample size, and the polling average. This is a simple exercise, and the weights for the regression-polling blend in particular might be off, but I think it’s useful as a rough guide to the election. Here’s what the current lay of the land looks like:

A few things immediately jump out. The Midwestern swing states are all light blue, for example. But those Arizona numbers sure are something. And look at Texas! The model thinks Biden is down by 2 points there. Georgia is also competitive. Is the sunbelt the new rustbelt?

Here’s another chart that shows how far to the left/right each state has moved since 2016. Let me explain what’s going on. In Arizona, for example, the chart shows that Biden’s polling relative to his national polling lead is about three and a half percentage points higher than was Clinton’s margin in Arizona relative to the national vote. Similarly, Texas has moved three points to the left.

All these numbers are particularly revealing, but let’s focus in on Texas now. The shift there is particularly notable in part because it moves the state nearly to the border between a red and blue state, and also because the Lone Star State has the second largest haul of electoral votes.

This means we should expect it to be important in determining the electoral college winner come November. We can quantify its influence with a value called a “tipping-point” index, which is the percent of the time we expect the electoral college winner to be decided by a particular state.

To determine these values for each state, I built a simple little toy election simulator that varies the polling average in each state by the historical distribution of national, regional and state-level polling errors at this point in the election cycle. You can find all the data and code for this analysis up on GitHub. Those tipping-point probabilities look like this:

The data show that, if the polls are right, Texas will pick the winner of the election about 12% of the time. That’s higher than for any state aside from Florida, which the model thinks will pick the winner in 14% of simulations.

This table is the most important takeaway, I think. That Texas is the second most important state in determining the winner this November is a pretty notable deviation from last cycle. But Georgia and Arizona are also both pretty high up there, which is worth noting.

If the polls are right, the sunbelt will the the place to watch this fall.


Posts for subscribers


What I'm Reading and Working On

I’ll be writing this week about which country’s leaders have seen the biggest increase in their approval ratings since the outbreak of covid-19, and about what this tells us about how people in different countries perceive and react to threats.


Thanks for reading!

Thanks for reading. I’ll be back in your inbox next Sunday. In the meantime, follow me online or reach out via email if you’d like to engage. I’d love to hear from you!

If you want more content, I publish subscribers-only posts on Substack 1-3 times each week. Sign up today for $5/month (or $50/year) by clicking on the following button. Even if you don't want the extra posts, the funds go toward supporting the time spent writing this free, weekly letter. Your support makes this all possible!