Texas Gubernatorial Project

Overview of the Experiment

The Texas Gubernatorial Project (TGP) commissioned Evolving Strategies to test the effectiveness of pro-life messages in shifting vote preferences in the Texas gubernatorial election.

ES executed a randomized-controlled experiment using an online survey sample of respondents matched to the Texas voter file the week of September 29th, 2014. 

Our target population was broadly representative of the registered voter population in Texas. We randomly assigned each voter to one of 4 conditions in a true experiment – 3 treatment groups and a control group (details on the experiment can be found in later sections). Just over 2,000 respondents completed the survey experiment, with about 500 in each condition. Hispanic voters comprised over 12 percent of the respondents.

All respondents took the same survey; the only difference between groups was whether the voter was exposed to one of three pro-life radio ads before answering vote preference and other survey questions (treatment groups) or did not view these before the questions (control group). Please see below for the exact wording of the radio ads.

Statistical analyses were then performed to identify significant treatment impacts on vote preference and other outcomes measured in the survey.

(See full report here.)

Use of the Experiment

The data from this experiment was used to model the expected impact of the ads on the vote probabilities for each of the approximately 13 million voters in Texas. 

In other words, it allowed us to identify the best targets for persuasion across all of Texas, based on the impact of the ads on the survey response and their demographic, consumer, and past voting data. We did not model what voters thought of the ads, but rather how the ad was likely to shift their probability of supporting Greg Abbott or Wendy Davis.

We then combined these modeled targeting data with the results of a previous, large scale field experiment using phone messaging to produce the best possible targets for specific messages to be delivered in the final days of the gubernatorial election.

These results were also used to guide the development and deployment of Hispanic-language radio ads on stations across Texas in the final weeks of the election, as well as online ads.

Overview of the Core Findings

The message experiment results indicate pro-life messages that attack the extremism of Wendy Davis have a generally positive impact, although there are significant differences in how voters respond based on gender, ethnicity and age in particular.

  1. Women respond particularly well, while we see significant pockets of backlash from white male respondents.
  2. Hispanic voters overall move significantly and very consistently away from Davis toward Abbott. Hispanic women respond best out of all subgroups analyzed. 
  3. The Standard and Hispanic Babies messages appear to outperform the Not Personal message. 
  4. Younger voters respond very well, the average impact decreasing with age.
  5. Weak Democrats respond best among partisan segments, followed by Weak Republicans and Strong Dems. The impact among Strong Republicans is much diminished, and in the case of male Strong Reps turns negative on average.
Evolving Strategies

Get in touch today