Abstract: SDAN volunteers sent 3 waves of educational postcards to low-mid turnout propensity Democratic-leaning voters in Virginia, Mississippi, and Louisiana in the first half of 2019 to determine if these early voter education efforts would boost turnout in the 2019 state legislative primary and general elections. We found that receiving postcards did not boost voter turnout in the primary or general election, and there was some marginally significant backlash to postcards in the primary election. Further, postcards that had general messages appear to have elicited more backlash than postcards that had more specific messaging, and even marginally significantly predicted a decrease in voting in the general election. More research is needed to contextualize these results.
Objective: This study sought to determine if sending 3 waves of handwritten postcards to high support, mid-low turnout propensity voters (i.e., people who support the Democratic party but need motivation to turn out to vote) could increase voter turnout for the 2019 Virginia, Louisiana, and Mississippi primary and general elections.
Background: Industry research suggests that sporadic voters are often ignored by political campaigns, which may see them as bad investments based on their likelihood of voting. This means that sporadic voters often do not receive as much attention, or as many asks, as more reliable voters.
However, we know that effects from persuasion and GOTV communications with voters generally “decay” and become weaker and/or disappear after a period of days or weeks. More research is required to see if this is true with low propensity/high support voters.
There are several unique circumstances related to these elections and this sample that may have contributed to an electoral context that is not generalizable. All of these elections were conducted in odd years, which means that the ballot does not include federal candidates (i.e., congressional, Senate, and Presidential candidates). Such elections tend to draw less interest, less media, and less money to the state, and voter turnout is generally much lower than it is in Congressional midterm or Presidential years. Further, these states are not representative of the larger American electorate and represent only one region of the US.
Additionally, due to the early timing of the first postcard delivery, the districts and targets were chosen before the candidate filing deadline in each state. This meant we chose to include some districts (including, unfortunately, most of the districts chosen in Virginia) that ended up not having state legislative primaries. Since Virginia did not have a Gubernatorial race like Mississippi and Louisiana, this meant that most of the Virginia sample did not vote in the primary. These considerations limit our ability to draw reliable conclusions about this voter education tactic from this particular study.
Specifics: SDAN pulled a list of all registered voters in 4 Virginia House of Delegates Districts, 6 Louisiana state House of Representatives districts, and 13 Mississippi state House of Representatives districts who met inclusion criteria (registered voters living in one of the above districts with partisanship scores between 80-100, indicating general support for Democrats, and turnout propensity scores for off-year general elections between 30-50, indicating sporadic/low voting behavior). The sampling frames for these states were: LA – 20,255 voters, MS – 17,010 voters, and VA – 18,610. Each list was randomized and up to 18,000 voters were selected from each list. For Louisiana and Virginia this resulted in samples of 18,000 from each state, and for Mississippi this resulted in a sample of 17,010. This resulted in a final sample of 53,010.
All participants were randomized into either the postcard condition, where they received 3 waves of postcards in the first half of 2019, or the control condition, where they received no postcards. Volunteers wrote to voters in one or more states. All recipients received the same first postcard, but for the second and third postcards, postcard recipients were randomly assigned to receive postcards with either a general or a specific message. The general message for postcard 2 outlined a few general issues that Republicans were failing to deliver on in the state, while the specific message cited specific examples. The general message for postcard 3 outlined a few general issues that Democrats plan to deliver on if given the majority in the state legislature, while the specific message cited specific examples. SDAN consulted with organizations and volunteers local to the state to generate appropriate state-specific messaging.
Volunteers mailed postcards to an in-state partner, who then mailed them locally. After the 2019 elections, SDAN matched all individuals in both conditions to the voter file to determine if they voted in the primary and general elections in 2019.
1. Postcards largely did not affect primary or general election voting.
While postcards (both message conditions combined) had a marginally significant negative effect on voter turnout in the primary election (p = 0.065), they did not approach significance for predicting voter turnout in the general election (p = 0.196).
- For the primary election, postcards were associated with an approximately 5.7% decrease in odds of turning out to vote compared to folks who did not receive postcards.
- For the general election, postcards had no effect on voting, indicating that any backlash during the primary from the postcards was short lived.
2. More specific messages are better than general messages.
In both the primary and general elections, it appears that the generally worded postcards were associated with a decrease in voter turnout, while the postcards that cited specific examples were not associated with a significant decrease in turnout.
- For the primary election, people who received a generally worded postcard were marginally significantly less likely to turnout to vote in the primary election than people who did not receive any postcards (controls; p = 0.054). People who received postcards with specific examples were not distinguishable from controls (p = 0.179), indicating that while they did not have a positive effect on turnout, they were not associated with the backlash that the general message was.
- There were similar trends for the general election, with marginally statistically significant backlash for voter turnout among generally worded postcard receivers (p = 0.062) and non-significant results for the postcards with specific examples (p = 0.698). This indicates that the general postcard may have actually had some longer term effects that were obscured when both postcard messages were combined in the previous model.
Caveats and Considerations
- Unique participants. Study participants came from selected competitive districts in 3 specific states (LA, MS, VA). Further, these voters were high support, low to mid turnout propensity voters, indicating they are a very specific population of people and the findings from this sample cannot be widely generalized.
- Quiet election. Odd-year election cycles are “quieter” than other types of election cycles as fewer people are running and the races are generally for state and municipal elections. This means that results from these elections are not generalizable to elections in other contexts.
- Statistically underpowered. This study was also underpowered to detect effects and would need to be replicated in a larger sample size to determine if the results we see are reliable.
Contributions and Future Directions
This is far from the definitive word on whether vote education postcarding can boost turnout, but results suggest that we should tread carefully in voter education efforts using postcarding, as they do appear to have minimal electoral consequences even months later. It is also possible that the effects the postcards have are detrimental, meaning that this technique should be studied further in order to determine what produced the backlash in this sample.
It’s important to note that the postcards sent in this study were delivered weeks and months before the elections we were trying to turn people out in. Because prior research has shown that persuasion effects often decay, our results are not unexpected. More research will need to determine how to overcome the ‘decay’ associated with earlier persuasion efforts like this, or, conversely, more research should be done using this voter education postcarding tactic closer to elections.
Future directions for this research should include finding other ways to effectively deliver voter education messages and testing these methods outside of the same states. Garnering more information on this technique will allow folks to make more informed choices about postcarding and messaging for these sporadic Democratic voters. Overall, the current evidence should be taken with caution as an initial suggestion that this type of messaging needs more work.
If you’re interested in reading more about this study, a longer report is here: Three Wave Voter Ed Postcard Short Report.
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 Mallory.