Abstract: This study tested the effect of receiving a text message about their local state house or senate candidate’s participation in the January 6, 2021 insurrection on constituents’ likelihood of voting in the November 2022 general election. The study design was a randomized controlled trial, with a control condition (in which voters received no contact) and a text message condition in which voters received an informational text message. The texts were sent to voters in 15 different states, in districts where insurrectionists were running for state legislative offices against a non-insurrectionist opponent or incumbent of the other party, ahead of the 2022 general election. Inclusion criteria were a partisanship score of 70+, a turnout score for midterm general elections of 0-80, and having a cell phone number listed in the voter file. Results suggest that the text messages had a negative but only marginally significant effect on voter turnout, compared to those who received no communication. In the future, this study could be replicated with a larger sample and possibly with changes to the treatment itself: In addition to the negative statement about the insurrectionist candidate, the message could also include a positive statement about the insurrectionist’s opponent, to potentially encourage voter turnout.

Objective: This study sought to determine if educating voters about state legislative candidate involvement in the January 6, 2021 insurrection at the US Capitol would increase voter turnout in the 2022 general election, in the districts where insurrectionist candidates were running.

Background: This study used a randomized controlled trial (RCT) designed to determine if accountability-based messaging around anti-democratic/aberrant social behavior would increase voter turnout among the constituents who were at risk of having an insurrectionist state representative or senator elected to represent their district.

Specifics: The RCT was designed by SDAN. The study targeted registered voters living in particular legislative districts where an insurrectionist was running for state legislative office and was being challenged by a candidate of the other party. The target districts were located in Alaska, Arizona, California, Colorado, Florida, Iowa, Michigan, New Mexico, North Carolina, North Dakota, Rhode Island, Texas, Vermont, West Virginia and Wisconsin. The sample was further narrowed to voters in these districts who met additional inclusion criteria—modeled midterm election turnout scores between 0-80, modeled partisanship scores of 70+ (i.e., likely Democrats), and cell phone numbers listed in the voter file. From this universe of voters, 70,000 were randomly assigned to the texting condition. Remaining voters were assigned to the control condition and received no contact. Because of deliverability issues, text messages were actually sent to only 69,072 of the voters assigned to the treatment condition. The final n for the study was 267,730. SDAN staff sent text messages to voters on October 26 and October 27, 2022. After the election, state turnout records were matched back to TargetSmart’s voter file to determine if the individuals targeted in this study voted in the 2022 general election. 

Basic Takeaways:

Receiving the insurrection text message did not increase the probability of voting. The turnout rate was 0.3 percentage points lower among people who received text messages compared to those who did not (the control condition), and this effect was marginally statistically significant (p = 0.08). 

The effect of the treatment on voter turnout did not vary significantly depending on the target’s predicted midterm turnout score, or the incumbency status of the insurrectionist candidate, or the insurrectionist’s gender. People did not respond differently to the treatment (receiving a text message) based on any of these potentially moderating factors.

This study had several limitations, including that it was statistically underpowered and used very specific inclusion criteria to define the sample. The results of this study are not widely generalizable.

Key Findings: 

People who received the text message were 0.3 percentage points less likely to vote than people who received no communication. 

• This difference was marginally statistically significant (p = 0.08). It means that the insurrection text message did not increase people’s probability of voting; in fact, those who received the insurrection text message actually turned out to vote at a lower rate than the people in the control condition.

People’s responses to the insurrection text message did not vary significantly based on their modeled midterm turnout scores (i.e., their predicted likelihood of voting in a midterm general election).

• There was no statistically significant interaction effect between turnout score and treatment condition (p = 0.72). 

The text-message treatment had a more negative effect on voting among people in districts where the insurrectionist was an incumbent representative, but there is no significant difference based on incumbency status. 

• The effect of receiving a text message is slightly negative among targets in districts with an incumbent representative compared to those with a non-incumbent candidate, but it is not statistically significant (p = 0.46).

People in districts where a female insurrectionist was running were slightly more likely to vote after receiving the text message, compared to people in districts where a male insurrectionist was running.

• The text-message treatment had a small positive effect on turnout among the targets whose insurrectionist representative (or candidate) was a woman, but this effect was not significant (p = 0.24). 

Caveats and Considerations:

Generalizability is limited. Though this study involved voters in 15 different states, it specifically targeted voters who lived in a district where a candidate running for their state house or senate seat attended the January 6 insurrection. It was a relatively small group of candidates representing an extremist viewpoint, and it may be that these districts (the candidates and the voters) are more extreme than the average state legislative district.

The insurrection was hopefully a rare occurrence. If such extreme events don’t recur, replicating this specific study would be difficult. However, the findings contribute to a larger literature about holding candidates and elected officials accountable.

Data quality: Phone records in the voter file are not 100% accurate. Sister District contacted the targets in this study using the cell phone numbers provided in the voter file. However, it is far from certain that these phone numbers were correct or up to date. It is highly likely that some portion of the voters in the treatment condition did not receive the text message due to inaccurate contact information.

Conclusions and Future Directions:

Overall, the results indicate that the insurrection text messages had a small, marginally significant and negative effect on voting: The turnout rate was lower among people who received text messages than among those who did not. There were no statistically significant interaction effects between treatment condition and turnout score, incumbency status of the candidate, or whether the candidate was a woman. 

One explanation for these results could be that the reminder of the insurrection was not motivating to people who were not already highly likely to turn out. It should also be noted that the actual 2022 turnout rate among this study’s sample, 34%, was much lower than expected, based on the average modeled midterm turnout score (47%). Despite their predicted scores, this group of targets simply had a low turnout rate overall, regardless of whether they were in the treatment or control condition.

This study has a number of limitations, including the narrow inclusion criteria, the fact that it was only run in a handful of districts with extreme candidates, and it invoked a specific event that is unlikely to happen again, at least in the same way it happened on 1/6/21. The current results suggest that this tactic only served to depress turnout rather than mobilizing voters, possibly because the message itself was purely “negative” (i.e., focused on the insurrectionist). Negative messages by themselves sometimes have de-mobilizing effects on voters. In future replications of this study, the treatment should also include a positive message or an encouragement to vote for the insurrectionist’s opponent.

If you’re interested in reading more about this study, a longer report is here.

SDAN’s commitment: It is SDAN’s intention to provide as much context as possible to allow for the nuanced interpretation of our data. SDAN’s convention is to contextualize effects by reporting p values, confidence intervals, and effect sizes for all models tested (these items may be in the longer report linked in the blog). Additionally, SDAN always differentiates between planned and exploratory analyses and a priori and post hoc tests, and reports the results of all planned analyses regardless of statistical significance. If you have questions about these findings please email Jillian.