Before Colin Allred’s campaign even got word of the late August poll, the news had already spread to social media. Democrats gleefully touted that the congressman, who represents a Dallas-based district in the U.S. House, was on his way to unseating Senator Ted Cruz. The survey, from the University of Houston and Texas Southern University, found that Cruz led Allred by roughly two percentage points (47 percent to 45), within the margin of error. The party quickly pounced on the news. Though Allred had not been publicly scheduled to speak at the Democratic National Convention, just minutes after the poll dropped, a new speaking lineup included his name. “On November 5,” an email sent that day from campaign manager Paige Hutchinson read, “Colin Allred is going to beat Ted Cruz and give Texas the leadership we need.”
This might seem like a familiar story. Palpable excitement among the Texas Democratic base in the lead-up to elections, followed by crushing defeat in November, is not new; in fact, it happens around this time every two years. The cycle goes like this: A somewhat high-profile Democrat (let’s say, Beto O’Rourke or MJ Hegar) receives a smattering of polls that suggests they could finally flip a statewide seat and end the Democrats’ three-decades-long losing streak. Then more polls come out showing that these races may be close, but they’re not that close. Finally, come Election Day, the Republican candidate ekes out a win. (On August 1, 2018, for example, Texas Lyceum released a poll showing the race between Cruz and O’Rourke, his then-challenger, in a dead heat. Cruz later won his reelection bid by roughly three percentage points.)
Right on cue, just a few weeks after August’s UH/TSU poll, a handful of surveys found Allred down by a greater margin—though most within single digits. He’s losing by ten percentage points according to ActiVote, four points according to Emerson College, five according to Morning Consult, seven according to Quantas Polls and News, and eight according to the Texas Politics Project at the University of Texas at Austin.
Given the gyrations in Cruz’s and Allred’s standing in the polls, and Texas Democrats’ history of not winning, can you trust surveys suggesting that the Dems might flip the Senate seat? Which polls are accurate, and how should you read them? We put together guidelines for how to temper or stoke your hopes or fears of a blue Texas.
- Don’t get too excited by outliers. The UH/TSU poll was particularly favorable to Allred. But avoid the temptation to overreact to individual polls; it’s more helpful to use polling averages and to look at trends in the results found over time. In this case, both the UH/TSU and the ActiVote polls stand out from the pack. Since the March primary elections, according to FiveThirtyEight’s polling average of the race, Cruz had led Allred by an average of nearly seven percentage points. Although a favorable poll can sometimes mean the start of a new trend, it is often a fluke.
- Pay attention to possible methodical issues. It’s impractical to survey the entire state, so pollsters select a sample of Texans. That sample should represent the state—but it does so only in theory. You should always look at how many people were polled and whether the demographics are a reflection of the Texas electorate. Is it weighted to properly represent voters without a college degree, who are less likely to respond to pollsters? Does it match Texas’s demographics, or does it specifically match the demographics of voters, who are disproportionately older and whiter?
The margin of error (sometimes labeled as “MOE” or a “confidence interval”) will increase as sample size gets smaller; between 1,000 to 1,400 respondents is considered good for a statewide survey. In this case, the Texas Politics Project and Quantas surveys seem to be the best polls—if we’re grading them on this metric alone. (In both survey methodologies, too, the pollsters noted that respondents were matched to a representative sample of Texas’s electorate, by age, education, gender, and race.) And unlike ActiVote, which surveyed only four hundred likely voters, both the Texas Politics Project and Quantas have smaller error margins. That’s the pollsters’ way of disclosing that their survey might not be exact but that it gets within a certain number of percentage points.
- Note what types of voters are being polled. Related to point two, polls of “likely voters”—registered voters who have indicated to surveyors that they have a strong intention to vote on Election Day—tend to be more accurate than polls of registered voters (which, in turn, tend to be more predictive than polls of voting-age citizens). Indeed, higher percentages of likely voters cast ballots in presidential elections than registered voters.
Early in an election cycle, when fewer folks have decided whether they intend to vote, most pollsters stick with samples based on registered voters. Closer to the start of early voting, usually after Labor Day, they switch to likely-voter surveys.
The Texas Politics Project surveyed registered voters, while the UH/TSU poll was conducted among likely voters. That doesn’t mean the Texas Politics Project poll isn’t a good survey; it has a strong record of accuracy and should still be considered in polling averages of the Senate race. But more weight should be placed on polls that survey likely voters, especially as we get closer to November.
- Consider the source. Often partisan groups—or even campaigns—will release their own polls that paint a rosier picture for their choice candidate; these should be treated with healthy skepticism. The six recent surveys of this race, however, were all conducted by nonpartisan outlets. Can you trust them?
At least two of the six pollsters have good track records: The Texas Politics Project, for one, correctly predicted in October 2020 that Senator John Cornyn would beat Hegar by eight percentage points. Similarly, in October 2018, Emerson had Cruz leading O’Rourke by a margin that nearly mirrored the junior senator’s Election Day results. Other pollsters—notably Quantas—are newer on the scene, which means you should probably treat their surveys with greater caution until we can assess whether they match up with the results.
- Look at the dates. Pollsters don’t release data the day it’s collected. So it’s possible that a “new” poll contains some old(ish) data. For example, Allred’s team jumped on the UH/TSU poll, which appeared to show a “Kamala Harris bounce” for Allred even though half of the survey was conducted before President Biden announced he wouldn’t seek a second term. Keeping in mind the gap between the surveying and the reporting of results is important. Pollsters recommend waiting a few days after these events and to heed polling data conducted after those dates.
At this point in the race, it’s important to remember that polls are snapshots, not predictions. What data we have paints a consistent but incomplete picture; in this case they’re finding that Cruz enjoys a single-digit lead over Allred. We can comfortably say the race is competitive, but Cruz is still a solid favorite.
To more fully see the picture of the race, we also looked at smaller-scale data not as readily available to the public. Many Democratic and Republican consultants are trying to gauge voter attitudes in more bite-size pieces, whether it’s for a state House race or to see if they can pick up on trends that might carry over into other competitive areas. We reached out to campaigns across Texas for their more granular surveys. Two internal polls shared with Texas Monthly, both conducted by Democratic groups, confirm the top line of the public, nonpartisan polls.
Take, for instance, the most recent survey commissioned by the Democratic Lone Star Project. The poll was taken before Kamala Harris became the Democratic presidential nominee, but it contains useful information on an important bellwether—Tarrant County, home to Fort Worth. Matt Angle, a veteran Democratic party strategist, surveyed more than 1,400 likely voters there and found Allred ahead of Cruz in the county by two percentage points (51 percent to 49 percent, which was within the survey’s margin of error).
Allred will need to improve his standing here slightly if he wants to beat Cruz. In general, Democrats need to run up the margins in big cities to counterbalance Republicans’ domination of the rural parts of the state, where in the last three elections the GOP has garnered no less than 70 percent of the vote. Fort Worth isn’t a Democratic stronghold like Dallas and Houston are, but Angle predicted that to win a statewide race, a Democrat would need to field at least 53 percent of the vote in Tarrant County. That would represent a high-water mark for Democrats in recent history: O’Rourke, in 2018, won 50 percent of the vote to Cruz’s 49 percent; while Hegar lost the county to Cornyn two years later by about five points (en route to falling short statewide by ten points). That same year, Biden narrowly won Tarrant County over Donald Trump as he suffered a statewide defeat of five percentage points.
We also have polling in the Dallas area that augurs somewhat unfavorably for Allred. Jeff Dalton, a Democratic campaign strategist, conducted a mid-August survey of likely general election voters living in Dallas and five surrounding counties—Collin, Grayson, Hunt, Kaufman, and Rockwall. He found Allred has an edge of nearly ten percentage points there—behind where he needs to be to win the whole state. In his 2018 loss, O’Rourke carried these six counties by roughly thirteen percentage points over Cruz; Biden, meanwhile, won these counties over Trump by eleven percentage points, Dalton says.
Of course, these are just a few data points and a lot can change before November; the only poll that matters, the idiom goes, is the tally on Election Day. For now, the best way to read polls is to look at the average consensus. The race is still very much Cruz’s to lose.
