Center for Opinion Research
The Department of Political Science & History operates a full scale polling center, which includes professional computer assisted telephone interviewing (CATI) software and 25 work stations for students to learn about survey research by experiencing how the data is collected.
The September survey by the Dallas Morning News and UT Tyler is the fifth poll that has studied the views of Texas voters during the 2020 Election. This survey took place immediately after the 2020 National Party Conventions. The survey also includes 1,176 registered voters to provide insights about the general election. Within the sample of registered voters were identified 901 likely voters by verifying voter histories and limiting our analysis to those who responded they were "extremely likely" to vote in November.
The Dallas Morning News/UT Tyler Poll reflects a statewide random sample of 1,176 registered voters during the six days between August 28 and September 2. The mixed mode sample includes 332 registered voters who were surveyed over the phone by the Center for Opinion Research with support from ReconMR and 844 registered voters that were randomly selected from Dynata’s panel of registered voters that was stratified to match the demographics of Texas’s registered voter population. The online and phone surveys were conducted in English and Spanish.
Post-stratification weights were calculated for the random telephone sample and web sample separately to be representative of the Texas registered voter population, before the two weighted samples were combined into one standardized sample (see Elliott 2009). To balance sample demographics with the estimated gender, age, race/ethnicity, and education of registered voters in the state we use an iterated process known as raking. These parameters were derived from 2018 Current Population Survey to reflect Texas’s electorate. The use of these weights in statistical analysis ensures that the characteristics of the sample closely reflect the characteristics of registered voters in Texas. The second step uses weights from the probability phone sample to standardize the weights for the non-probability online sample to reduce sampling bias (see Elliott 2009).
In this poll, the sampling error for 1,176 registered voters in Texas is +/- 2.87 percentage points at a 95 percent confidence interval. Within the sample there are also 901 likely voters, who are “extremely likely” to vote in November’s election. Also, in our phone sample we verified how frequently the vote participated in past general elections. The margin of error for our sample of likely voters is 3.27.