This is the second of two analyses that will attempt to prognosticate Tuesday’s general election. Here we focus on individual races. Our previous analysis examined the state’s political climate entering the election.

Borrowing from Monty Python, “That’s no ordinary gubernatorial election.” Early voting much more closely resembled a presidential election, and more early votes were cast than total votes in 2014, including Election Day. So we can toss gubernatorial election year assumptions out the proverbial window. Unfortunately, almost all of the polling done to date was based on turnout assumptions that we just tossed out the window.

Yet, it isn’t a presidential election, either. Turnout resembled the 2016 early turnout but did not match it. Every new voter that cast a ballot in the last two weeks meant one more 2016 early voter did not, and there’s reason to believe that some of those were Democrats. This year, turnout for some blocs critical to Democratic success lags behind key Republican blocs, so far. For example, we dove into precinct-level early voting turnout for SD10 precincts where at least 500 votes had been cast. In precincts where at least 75% of straight-party votes were cast by Republicans in 2016, early voting turnout averages 51% of registered voters. In precincts that were at least 75% Democratic, turnout averages 30%.

The challenge is trying to model how much this election will behave like a presidential election and how much it will behave like a gubernatorial election. The other challenge is trying to figure out what we don’t know, which is, of course, more that we would like to admit. We have run the numbers and made the most educated assumptions and decisions we can, but our predictions are best thought of as a preseason Top 25. They have to play the games before you know how right, and how wrong, those predictions were.

Summary of Predictions

We predict the following outcomes:

  • Republicans sweep the statewide races, including the U.S. Senate race, which we project U.S. Sen. Ted Cruz (R) will win by 5 points, and the governor’s race, which we project Gov. Greg Abbott (R) will win by 17 points
  • The average statewide Democratic candidate will receive 44.4% head-to-head against the average statewide Republican candidate
  • HD47 (Workman), HD105 (R. Anderson), HD114 open, HD115 (Rinaldi), HD136 (Dale) and CD7 (Culberson) will flip to Democrat from Republican
  • No Democratic seats will flip to Republican; and
  • Every other seat will remain in the hands of the party who currently holds it.

That said, if the vote breaks the right way in the right places, we would not be surprised to see one or more of SD16, open HD52, HD108, open HD113, HD134, CD23 or CD32 flip to the Democrats.


Our model “floats” districts upon a “sea level” of the estimated average statewide Democratic nominee’s vote percentage calculated head-to-head against the Republican nominee. Yesterday, we set the “sea level” at 44.4%. Thus, a neutral Democratic nominee can be expected to defeat a neutral Republican nominee if the district is at least 5.7% bluer than the state as a whole. A Democratic nominee who overperforms the statewide slate can prevail in a district that is not quite as blue, and one that underperforms would need the district to be even bluer.

So the key calculations are estimating the district’s float relative to the sea level of the state and estimating the over- or underperformance of each candidate relative to their respective party’s statewide slates. We modeled the float based upon trends in gubernatorial election cycles (1/3 weight) and presidential election cycles (2/3 weight). Think of it as extending the historical trend line into this year. A district that has been getting bluer, relative to the state, would be estimated to get a little bluer.

Analyzing a district relative to the state allows us to control for variations in the statewide candidates’ performances from year to year. As such, it’s possible for a district to get “bluer” even as the Democrats’ performance in the district nominally worsens. For example, the statewide Democratic candidates averaged 41.5% of the vote in SD16 in 2012 and 38.5% in 2014. Yet, the district got 0.9% more Democratic, relative to the state, going from 0.8% redder than the state as a whole to 0.1% bluer. This creates far clearer trends at the district level than just raw vote totals.

So far, we’re looking only at neutral candidates, which means they are indistinguishable from the average statewide candidate. Most incumbents have a history of how they perform relative to the statewide slate, so that adjustment can be made to the district’s “float,” but few challengers do. This year we emphasized fundraising (relative to the opponent), based on the knowledge that virtually all candidates who flipped seats in recent years were financially competitive, and gender, based on the assumption, derived from the gender gap observed in polling, that a woman candidate will get some sort of boost when running against a male candidate.

The model yields nine Republican-held seats – one Congressional, one Senate and seven House – that would flip if all the assumptions are correct. It is more bullish on some seats (HD45, HD96, SD10 and CD23) than I am, and it is more bearish on some seats (HD114, SD16, CD7 and CD32).

Toss Ups

The model puts four of our “toss up” races at 50% Democratic or higher (measured head-to-head against the Republican):

  • HD105, Rep. Rodney Anderson (R-Grand Prairie) versus Terry Meza (D), a rematch from 2016
  • HD115, Rep. Matt Rinaldi (R-Irving) versus Julie Johnson (D)
  • HD134, Rep. Sarah Davis (R-Houston) versus Allison Sawyer (D); and
  • HD136, Rep. Tony Dale (R-Cedar Park) versus John Bucy, a rematch from 2014.

We expect HD105 and HD115 to flip. Both districts were won by Hillary Clinton in 2016 and saw a Democratic advantage in straight-party voting. Both Democratic challengers are women running against two incumbent men. We looked at nine individual metrics to evaluate the potential of a seat to flip, and HD115 ranked in the top seven in all but one. It ranked first in the shift of straight-party vote advantage toward the Democrats, and it was ranked highest among House races in terms of the challenger’s fundraising. The only two metrics going in Anderson’s favor are the fact that he overcame the largest straight-party deficit of any candidate to do so in 2016 and his challenger’s fundraising is below the threshold we expect for a typical successful challenger.

In HD134, we expect Davis will win enough support from the full-ballot voters to turn away her underfunded challenger, but Davis could be swamped by straight-ticket voters if they swing significantly more Democratic. Her opponent is far behind in fundraising. Had Davis lost her primary race to a more conservative candidate, this district would flip. HD136 is more of a wildcard. Dale outperformed the Republican statewide slate by 1.4% in 2016 but was level with it in 2014, a race that involved his current challenger. Between 2002 and 2014, the precincts currently comprising HD136 got 14.4% bluer, relative to the state, the largest shift for any Republican-held district. We expect that trend to continue and predict HD136 will flip.

The model puts three of our “toss up” races at below 50% Democratic:

  • HD114 open, Lisa Luby Ryan (R) versus John Turner (D)
  • CD7, U.S. Rep. John Culberson (R-Houston) versus Lizzie Pannill Fletcher (D); and
  • CD32, U.S. Rep. Pete Sessions (R-Dallas) versus Colin Allred (D).

Hillary Clinton carried HD114 by 9.3% over Donald Trump, her third largest margin in a Republican-held district. The district is trending bluer and was 3.5% bluer than the state as a whole in 2016. Rep. Jason Villalba (R-Dallas) outperformed the Republican statewide slate by 5.4% that year, but he lost the primary to Ryan. The model does not translate Villalba’s past performance to her. If this was “The Price Is Right,” the model pegs HD114 as the closest without going over at less than half a point below 50%, so any tiny swing in one of our assumptions could put the seat over 50%. The race is a test of just how blue Dallas Co. is getting, and we believe the model is slightly understating that potential. We predict Turner will win narrowly over Ryan.

CD7 is less than 2 points below 50% in the model. The district was 2.9% redder than the state as a whole in 2014 but 1.3% bluer in 2016. It was 14.9% and 14.4% redder than the state in 2002 and 2004, respectively, which means it has moved the furthest blue of any Republican-held congressional district. Culberson has run about even with the average Republican statewide candidate for the last three cycles, but he underperformed it by 2.6% and 2.0% in 2006 and 2008, the last years he faced a (semi-)serious challenge. The model only considers the most recent elections, but an incumbent with his seniority should perform better than Court of Criminal Appeals candidates. We overrule the model and predict Fletcher will eke out a narrow win.

The model puts CD32 behind several other seats we consider far less likely to flip, such as HD96 (Zedler) and HD97 (Krause), a pair of Tarrant Co. seats we rated “likely Republican.” CD32 is ever so slightly bluer than the state as a whole, but the needle has not moved much in recent years. Unlike Culberson, Sessions has outperformed the Republican statewide slate in all but two years since his election in 2002. In one of those years, he was unopposed. That year was 2016, so the model awards him less credit than he deserves. Allred has out-raised Sessions, and a just-completed New York Times/Siena Coll. poll shows him leading the incumbent by 4, but we’re not convinced. We predict Sessions will narrowly retain the seat, but, if Dallas Co. goes bluer than expected as a whole, we would not be surprised to see an Allred victory.

Lean Republican

The model puts four of our “lean Republican” races and one “likely Republican” race at 50% Democratic or higher:

  • SD10, Sen. Konni Burton (R-Colleyville) versus Beverly Powell (D)
  • HD45 open, Ken Strange (R) versus Erin Zwiener (D) – the likely Republican one
  • HD47, Rep. Paul Workman (R-Austin) versus Vikki Goodwin (D)
  • HD113 open, Jonathan Boos (R) versus Rhetta Bowers (D); and
  • CD23, U.S. Rep. Will Hurd (R-San Antonio) versus Gina Ortiz Jones (D)

When she was elected to represent SD10 in 2014, Burton underperformed the Republican statewide slate by 1.5 points. We have previously explored how elections for the seat might be different if they were held in presidential election years. While this election looks more like a presidential election than a gubernatorial election, the turnout pattern we mentioned above suggests Burton has the advantage despite our model’s results. We predict Burton will narrowly win re-election.

In HD47, Goodwin has raised 44% of the amount raised by Workman, which is what we previously defined as the floor for a successful flip. Our model gives Goodwin a 1.0% boost for being a woman running against an incumbent male, and that boost is what puts the district, barely, over 50% Democratic. It would give her a bigger boost if her fundraising were more competitive. Workman has underperformed the Republican statewide slate in two of the three contested races he’s run. In a very high turnout race, we project a narrow Goodwin win, especially if a significant number of Republican-leaning women vote for her over him.

HD113 is another district where more Democrats voted a straight-party ballot than Republicans. Rep. Cindy Burkett (R-Sunnyvale) outperformed the Republican statewide slate by 3.7% in 2016 after outperforming it by 2.0% two years earlier. The model gives no such advantage to Boos, a considerably more conservative nominee. Bowers is given the same 1.0% boost as Goodwin in HD47, but her fundraising performance is worse relative to Boos. Bowers underperformed the Democratic statewide slate by those 3.7% two years ago. We expect Boos to win.

Finally, CD23 by all rights and reasons should flip in this political climate, just as it should have in 2016. Jones has out-raised the incumbent in a district that has a built-in Democratic straight-party advantage in presidential election years. In a typical gubernatorial election year, Republicans have a smaller straight-party advantage, but Jones should be able to overcome it. Yet, we are skeptical about Hispanic/Latino Democratic turnout, particularly in light of the SD19 special and runoff elections and the relatively low turnout in Hispanic/Latino-majority counties we tracked during early voting. We expect Hurd to be re-elected.

The model puts one of our “likely Republican” seats at 50% Democratic: open HD45. Former Rep. Patrick Rose (D-Dripping Springs) routinely outperformed the Democratic statewide slate by 10% or more until his 2010 defeat to current Rep. Jason Isaac (R-Dripping Springs). Rose outperformed the Democratic slate by just 7% that year. The district, as it is configured today, is bluer than the state as a whole and has been since 2004, but it isn’t trending in any direction. HD45 was 2.8% bluer than the state as a whole in 2016, the same percent as in 2008. The fact that the seat is open and the Democrat is a woman running against a Republican man caused the model to tip the seat into a flip. We disagree, in no small part because of Erin Zwiener’s (D) fundraising disadvantage, but we move it to the “lean Republican” from “likely Republican” column.

The model puts the remaining races rated “lean Republican” below 50% Democratic:

  • SD16, Sen. Don Huffines (R-Dallas) versus Nathan Johnson (D)
  • HD52 open, Cynthia Flores (R) versus James Talarico (D)
  • HD102, Rep. Linda Koop (R) versus Ana-Maria Ramos (D)
  • HD108, Rep. Morgan Meyer (R-Dallas) versus Joanna Cattanach (D)
  • HD112, Rep. Angie Chen Button (R-Garland) versus Brandy Chambers (D)
  • HD138, Rep. Dwayne Bohac (R-Houston) versus Adam Milasincic (D)
  • CD22, U.S. Rep. Pete Olson (R-Sugar Land) versus Sri Preston Kulkarni (D); and
  • CD31, U.S. Rep. John Carter (R-Round Rock) versus M.J. Hegar (D).

The closest of these to 50% is HD108, and it is only under 50% because the model gives Cattanach a smaller boost than some other Democratic women challenging male incumbents. Some of the district’s Republican voters can be a bit squishy. Clinton received a majority vote in the district, and a decline in the number of straight-party Republican votes represented 18% of the overall shift in straight-party advantage toward the Democrats. Republicans still had the advantage in 2016, but it was half of its 2012 size. Expect it to be close, but we expect Meyer to win, narrowly, unless it’s an even stronger night for Dallas Co. Democrats than we estimate.

The second closest of these districts to 50% is open HD52, where the model predicts the district will not be quite enough bluer than the state as a whole for it to flip. We are unable to evaluate whether either candidate will perform better than their respective parties’ statewide slates, and the model gives neither candidate any advantage beyond the projected partisan lean. If Flores were the Democrat and Talarico the Republican, then the model would put this district over 50% Democratic. We expect Flores to win, narrowly, but would not be surprised if it flipped.

Up next is SD16, for which the model has a bit of a blind spot because Huffines was unopposed in the 2014 general election. We have no sense as to whether he will outperform the Republican statewide slate, be level with it or underperform it. The district’s projected partisan lean, relative to the state, is not enough to flip the seat, but a very strong showing by Dallas Democrats could make the difference. We expect Huffines to prevail.

In HD102, Koop outperformed the Republican statewide slate by 3.3% in 2016, which offsets the blueward shift of the district calculated by the model. Her opponent’s fundraising does not meet the threshold we consider competitive, so her chances of victory depend on an influx of straight-party Democratic voters. Despite the seat’s very high rankings in many of the metrics we explored, we predict Koop will be re-elected.

The model puts HD138, CD22 and CD31 under 45%. All three start from a position of being redder than the state as a whole, and the model projects them to remain so despite their blueward shifts. HD138 has the potential to be an Election Night surprise, the sort of upset no one sees coming, as it has gone from 12.7% redder than the state in 2002 to 2.6% bluer in 2016. When he’s been opposed, Bohac has outperformed the Republican statewide slate, but he had no opponent last year. Milasincic’s fundraising total is below the threshold we consider competitive at this stage in the game.

The model’s second blind spot is CD22, where Kulkarni is trying to harness a polyethnic coalition that either generally votes Republican or doesn’t vote at all to boost him to victory over Olson. Fort Bend Co. is rapidly growing and diversifying, but it also historically has one of the highest rates of straight-party voting. More than three quarters of all ballots cast were straight-party votes. Democrats had the advantage in 2016, but the strongest Democratic precincts were in adjacent CD9, held by U.S. Rep. Al Green (D-Houston).

CD31 is a test of whether a candidate can transcend party identification. Hegar is the fundraising leader among challengers, more than doubling up Carter, and her advertisements have given her a national following. CD31 is moving blueward but is still red. A congressional Democratic candidate has not received more than 40% of the vote in these precincts since at least 2002. We predict Bohac, Olson and Carter will win, all by closer margins than most would expect.

The farthest district from 50% on this list is HD112. The district is only slightly bluer than the state as a whole, and it’s not enough to overcome Button’s ability to outperform other Republicans on the ballot. She will win comfortably, even if stronger than expected Democratic turnout occurs.

Lean Democrat

The model shows both of our “lean Democrat” seats remaining in Democratic hands. HD107, where Rep. Victoria Neave (D-Dallas) is seeking a second term against conservative Republican nominee Deanna Metzger, is expected to be closest race of any Democrat-held seat in the state, but we expect Neave to prevail by a few points. In HD144, higher-than-usual turnout is expected to keep Rep. Mary Ann Perez (D-Houston) from suffering the same upset she experienced in 2014. We predict Neave and Perez will be re-elected.

Other Districts

We predict all other seats to remain in the hands of the party holding them now.

The only possible exception to that is open ED12, where Democratic nominee Suzanne Smith has out-raised and outspent Republican nominee Pam Little for a Republican-held seat on the State Board of Education. We have not looked very closely at the district aside from calculating it was 3.5% redder than the state as a whole in 2016. Absent other analysis, we predict Little will hold the seat.

Other possible “sleepers” include HD129 (Paul), HD132 (Schofield) and HD135 (Elkins) in Harris Co. and HD65 (Simmons) in Denton Co. We have all of these rated as “likely Republican” and we expect the Republicans to win, perhaps by eye-raising thin margins.

Our live coverage of election results begins at 7 p.m. CST.