As I mentioned 4th of July weekend, I'm back. One of the posts I've long wanted to write, and caused me a lot of heartache in trying to put it together (due to getting the data), was the follow up to my last full piece from December 2020, a post-2020 election debrief.

Let's quickly recap that piece.

Polling and election forecast modeling are still useful

In both the 2016 and 2020 elections, Donald Trump ended up winning a higher proportion of votes than most pollsters predicted he would win. For many, this is a sign that polling is fundamentally broken. I strongly disagree.

It's beyond the scope of this piece to rehash the discussion, but if you want to read about what happened in 2016 polling Nate Silver's 12-part series is a good place to start. If you want to get into the weeds of polling methodology in 2020, Nate Silver wrote his usual post-election poll assessment this last March.

The long and short: While it's true that Presidential polls had fairly large errors in both 2016 and 2020, these poll errors overall were within historical ranges, and at least in 2020, the polls did correctly pick the ultimate winner (Joe Biden).

Relatedly, election forecast modeling is still useful. I personally like the FiveThirtyEight model as I've watched it grow and mature since 2008. Also, the team at FiveThirty Eight check the model after every election cycle to see how it actually worked. For the 2020 cycle, the model was well-calibrated - which means that events it predicted would happen 1/2 or 2/3rd of the time actually happened 1/2 or 2/3rd of the time.

Why is it important to care about polls? Because polls help those of us interested in changing the world understand, in a rigorous way, who cares about what and by how much. Without that information, we're just fumbling around in the dark.

The Racial Divide Drops and the BA Divide Grows

Despite what you might have heard, the racial vote divide actually dropped in 2016 relative to 2012 (source: NYT). That's because a large chunk of European/White American voters who had previously voted for Mitt Romney voted for Hillary Clinton. At the same time, some non-European/non-White American (i.e. Black American and Latin American) voters that had previously voted for Barack Obama voted for Donald Trump.

What was special about these voters? It turns out these particular Romney/Clinton voters often attended (and graduated) college, while the particular Obama/Trump voters often didn't attend college. So the Democratic Party ended up winning more European/White American votes while the Republican Party ended up winning more African American and Latin American votes - even though neither side won a majority of those respective groups.

And the same thing happened in 2020 (source: Pew Research, Catalist). Trump earned a higher proportion of Latin American votes in 2020 than he did in 2016, while Biden earned a smaller proportion of the same group than Clinton did. There was a much smaller, but still positive shift among African American voters. Similarly, Trump lost support among European/White Americans.

Source: Pew Research 

At the same time, college graduates continue to move towards the Democratic Party while non-college graduates continue to move away.

Source: Pew Research 

If you recall from last December, nationally, only 30% of adults 25 and older have at least a bachelor's degree (BA+) and the state with the highest percentage of BA+ (41%) is Massachusetts. If Joe Biden only won the 19 states (and DC) with more college graduates than the national average, he'd have lost the election.

As a reminder, here's what that map would have looked like.

Unsurprisingly, this is not too far off from the actual 2016 election map:

  • Swap Utah for Nevada (net-zero impact to the Electoral College split)
  • Swap Kansas for New Mexico (-1 from the Democrats)
  • Flip three of the four votes in Maine (+3 to the Democrats)

In other words, being the party of the college educated is not a great strategy for winning the Electoral College (or relatedly, the US Senate).

Ok, now that we've recapped, let's talk about what this means for Texas.


In blog speak, "tl;dr" stands for "too long; didn't read" and is supposed to just give the highlights. After writing this post, I realized it was quite long so I'll go through the new conclusions here, though you're welcome to continue reading if you want to go through the nitty gritty.

  • Texas State House Democrats thought we could pick up 9 seats (mostly in the suburbs) to build a new State House majority. In the end, we netted 0 seats, remaining at 67.
  • On the other hand, a majority of voters chose Biden for President in 74 House districts. Clearly there's some type of split ticketing going on.
  • Maybe it was just about Trump (though MJ Hegar running for US Senate also carried more districts (68) than State House Democrats did). Maybe those Republican incumbents were really beloved (in the nine Biden / State House Republican contests, all nine State House Republicans earned more votes than Trump).
  • At the very least, I think we can say six seats (HD-26, HD-66, HD-67, HD-108, HD-112, HD-138) had a fair number of split ticket voters. Had these six seats swung the other way, the Medicaid expansion vote would have flipped from 68-80 opposed to 74-74 tied.
  • Lastly there are some impressive Democrats in the Valley. Trump and Sen. John Cornyn won in HD-31 and HD-74, while Reps. Ryan Guillen and Eddie Morales, Jr., both Democrats, still carried the districts. Redistricting is not going to be kind to them, and we'll be in big trouble if we lose them next year.

Why State Legislative Seats Matter

While the Presidential election was certainly very important, I was actually more focused on State Legislature, which is now in the national news.

At the time I'm writing this,  Texas State House Democrats have broken quorum in order to block an elections bill that would among other changes, ban drive-thru voting, ban 24-hour voting, and grant partisan poll watchers more latitude in monitoring elections, potentially allowing them to learn how a voter might be voting.

For those that don't know, "breaking quorum" is a legislative tactic used to stop all business in a legislative chamber. Here's how it works. According to the Texas Constitution, the State House and State Senate are each required to have 2/3rd of their Members in attendance in order to conduct any business. So that means at least 100 members in the State House and 21 in the State Senate. So if 51 members of the State House or 11 members of the State Senate don't show up, those bodies can't conduct any business.

While there are many factors that go into deciding to break quorum, for most of the last decade, Democrats just haven't had the numbers because usually 10% of Democrats (often in the swingiest of seats) aren't willing to break quroum (for fear of losing re-election).

Year House Split (D-R) Senate Split (D-R)
2009 73-77 12-19
2011 49-101 12-19
2013 55-95 12-19
2015 52-98 11-20
2017 55-95 11-20
2019 67-83 12-19
2021 67-83 12-19

So big takeaway: every seat counts, even if you don't have a majority.

If State House Democrats hadn't won these 12 seats in 2018, we would not have the numbers to break quorum today.

Source: Texas State House of Representatives

An interesting note: seven of these 12 seats also voted at the statewide level for Greg Abbott and Beto O'Rourke, showing that some Texas voters will still split their ballots. A sometimes frustrating, but healthy sign of democracy!

Surprising Split Tickets in 2020

Ok, so what does this all have to do with the 2020 results?

Well, after flipping 12 seats in 2018, Texas Democrats wanted to shoot for a majority in the State House. They would need to flip 9 seats. They planned to target any State House seat held by a Republican that O'Rourke had lost by 10 points or less in 2018, while also holding onto the 12 seats won in 2018. Below is that list of 18 seats.

Source: Texas State House of Representatives

It was a pretty ambitious list, but also pretty intuitive. 11 of the seats were in the major urban counties: Harris (3), Dallas (2), Tarrant (5), and Bexar (1); and an additional five seats were in the major suburban counties: Fort Bend (2), Collin (2) and Denton (1).

In 2020, money was no object, Democrats had the advantage of a very unpopular Republican President at the top of the ticket, and voters were going through a once in a hundred year pandemic.

What was the result?

Texas Democrats picked up one targeted seat in Houston (HD-134) and lost one of our Houston pickups from 2018 (HD-132), leaving the partisan split exactly the same from 2018, 67-83.

But here's something strange: Joe Biden won 74 State House seats. Of the targeted seats, Joe Biden won nine. That seven seat difference is critical. If we had won those seats, Texas Democrats would have a majority in the State House right now (because two House Democrats won in seats that voted for Donald Trump).

Source: Texas State House of Representatives

I'm not entirely sure why Joe Biden's popularity at the top of the ticket didn't carry further down in Texas. FiveThirtyEight looked at split ticket voting across the country in 2020 and concluded "split-ticket voting was minimal in most [Congressional] districts". I'm not sure why it would be any different further down or maybe why Texas might be different.

But there is evidence of some split ticket voting in at least six districts (HD-26, HD-66, HD-67, HD-108, HD-112, HD-138). How can we tell? Well, in these six districts, the State House Republican earned more votes than Trump, and the Presidential contest had a higher number of voters than the State House race.

Source: Texas State House of Representatives

Maybe it's the case that in those districts the Democratic drop off (i.e. people who voted for Biden and left) was larger than the Republican drop on (i.e. people didn't vote for Trump, but stayed for the State Rep), which would explain why the turnout for the State Rep contest was lower than for the Presidential. But I think it's more plausible that a non-trivial number of voters split their tickets - especially since voters in these districts were much more likely to have gone to college (50%) than the state as a whole (30%).

That's incredibly important because had these six districts flipped, Democrats would have had 74 votes for the Medicaid expansion amendment, along with Rep. Lyle Larson (R-San Antonio), would have brought it up to 75, one shy of a majority.

Strong Democrats in the Valley

The other area where split tickets mattered was in the Valley. Two Democrats, Reps. Ryan Guillen (HD-31) and Eddie Morales, Jr. (HD-74) won their seats at the same time as voters in their districts chose Trump and Cornyn. That Trump and Cornyn were able to get support in those districts is surprising because in 2018, O'Rourke won both seats and Lupe Valdez (Democrat for Governor) won in HD-74.

This is part of that shift among Latin Americans voters towards the Republican Party that Texas Democrats need to be concerned about. Next week I'll go through the the Texas Democratic Party's Report identifying this trend and offer my own observations.

No matter what the cause, Reps. Guillen and Morales are now on the Associated Republicans of Texas and Hispanic Voter Network target list. With redistricting coming up, we really can't afford to lose those seats - because every seat matters.

I know this was a long post and I promise future posts will not be as long. But thanks for reading and please, send feedback to [email protected].