What Encourages Returning Citizens to Vote? Measuring the Impact of Different Forms of Voter Outreach in Iowa

On August 5, 2020, Governor Kim Reynolds issued Executive Order 7. This measure fully restored voting rights to Iowans with past convictions once the terms of incarceration, probation, parole, or special sentence were completed. The order was retroactive, applying to those who had already been released from the Iowa Department of Corrections. The order did exclude, however, individuals convicted of felony homicide offenses.

Despite this recent change in law, the political and legal landscape have fomented an air of uncertainty for many potential beneficiaries. Accordingly, voter turnout among this population can likely be bolstered by targeted information campaigns.

This paper aims to measure the impact of such an outreach effort. Specifically, we seek to measure how outreach using traditional mailers – a commonly used but expensive method for voter outreach – compares with outreach using social media and mixed social media/opt-in SMS campaigns. These treatments were chosen based on the feedback and past experience of our partner, the Campaign Legal Center, which has significant experience in assisting returning citizens regain voting rights.

See attached pre-analysis plan.

2022-10-10

2022-11-08

The attached pre-analysis plan provides more details. Brief overview:
--Differences in voter registration rates across different treatments in 2022 general election, as compared to one another as well as a pure control
--Differences in turnout rates across treatment groups in 2022 general election, as compared to one another as well as a pure control
--We will also analyze (most likely to be included in a subsequent work product) the long-run impacts of our intervention on future elections, particularly the 2024 primary and general elections.

We have obtained a full list of individuals who benefited from the change in law in Iowa in 2020. We have split these participants into two experimental arms, via a stratified randomization strategy, as described in our pre-analysis plan. We are scheduled to send out notifications to these individuals by early to mid-October 2022. Finally, we will collect data on voter turnout and voter registration as soon as it is available, likely before January 2023. Analyses will be conducted sometime between that time and March 2023.

In addition, we will collect data on voter turnout and voter registration for future elections in Iowa, in particular the 2024 primary and general elections. Depending on timing, analysis of such results might be included in the same paper as that which analyzes the 2022 election, or it might be included in a future paper; alternatively, it might be simply posted in a forum online.

This project will be conducted in accordance with the best practices to date associated with randomized-control trials. It will contribute to a growing corpus of research that aims at uncovering the impact of voter restoration programs and advocacy efforts in the same space.

To ensure that treatment arms are properly balanced across observable characteristics, we implemented a blocked or stratified randomization procedure. Operationally, we use the Python package stochatreat that allows for stratified randomization of any number of treatment arms.

Individual, stratified randomization procedure. We block randomize on the following:
1. Race/ethnicity inferred + true values (indicator variables)
2. Whether race/ethnicity was inferred by us (indicator variable)
3. Age Group (indicator variables)
4. Gender (binary variable)
5. Rights Ever Reinstated (binary variable)
6. Identified by Campaign Legal Center as a potential beneficiary (binary variable)
7. Pre-treatment voter registration status (binary variable)
8. Registered before Reynolds’s Executive Order (binary variable)
9. Address on File (binary variable)
10. Scott County Resident (binary variable)
11. Counterfactual, Non-Returning Citizen Available (binary variable)

No

N/A

We have a list of 132,118 eligible participants.

Arm 1 --Main (people with past felony convictions with known addresses): 15,983 (control), 15,990 (mailer), 16,004 (Meta, i.e., Facebook and Instagram, and Text Opt-in), 16,017 (Meta)
Arm 1 --Placebo (people without past felony convictions): 15,967 (control), 15,974 (mailer), 15,988 (Meta, i.e., Facebook and Instagram, and Text Opt-in), 16,001 (Meta)
Arm 2 (people with past felony convictions without known addresses): 22,705 (control), 22,701 (Meta and Text Opt-in), 22,718 (Meta)

See attached pre-analysis plan.