Voting and Motivated Beliefs

We will recruit eligible voters for the 2022 midterm elections in the US and present them with information about wait times. We will present selective but accurate information about wait times from the 2016 presidential election and randomly vary whether the subject learns about the 10% longest or 10% shortest wait time in their county. We will further contrast that information with the average wait time in the subject's state.

2022-11-01

2022-11-07

Perception Measures:
1. Perceived importance of midterm elections (1-7)
2. Perceived knowledgeability about the midterm elections (1-7)
3. Attitude to voting – duty vs. choice (binary)
4. Perception to influence politics by voting (1-7)
5. Perceived closeness of election race for seat in house of representatives (1-7)
6. Perceived closeness of election race for senate seat (1-7)
7. Perceived closeness of election race for governor position (1-7)
For items that are similar (e.g., closeness of race) we will create an index (e.g., by using PCA or z-scores).

Behavioral Measure:
- Propensity to click on a link that provides more information about poll worker jobs and allows to sign up for a job.

- Reported likelihood to vote
- Reported likelihood to vote conditional on bad weather, busy week
- Expected wait times in midterm elections

We recruit participants from an online platform (MTurk, Lucid or Prolific) and ask them to participate in a survey that asks for their views on elections and provides information about the 2022 midterm elections.

All participants are randomly assigned to one of two treatments. The treatments vary in what information about wait times is provided. Participants either receive information about the 10% longest wait times in their county from the 2016 presidential election or they receive information about the 10% shortest wait times in their county.
Based on that information, we study whether voters manipulate their beliefs about elections.

We plan to recruit 2,000 participants from an online platform (MTurk, Lucid or Prolific) to participate in a 10 minute survey. Before we provide our treatment intervention, all subjects have to answer some filter questions based on which we exclude certain participants (see below). Then, participants will be randomly assigned to either the Long or the Short treatment condition and receive accurate information about wait times in the 2016 presidential election. After the information provision, we elicit our primary and secondary outcome variables. The wait time data we use to provide the information stems from the study “Racial Disparities in Voting Wait Times: Evidence from Smartphone Data” by Chen et al. (2020).
The survey will launch 7 days prior to election and will continue until a day before election day. We will only survey participants who live in counties for which we can provide information about wait times.

Subjects will be recruited on from an online platform (MTurk, Lucid or Prolific) and will participate on a “first come, first serve” basis. Once they pass our selection criteria, they will be randomly assigned to our two (or three) treatment conditions.

Randomization will be done at the individual level.

No

Each observation is independent of the other observations. Thus, there is no need for clustering.

We plan to recruit 2,000 participants, 1,000 per treatment. Note that 1,000 subjects per conditions is a target value. Since it is difficult to determine ex-ante how many participants we can recruit who fulfill our participation criteria, the final number of participants might be lower. Exclusion criteria Our participants need to fulfill the following requirements: 1. Participants must be at least 21 years old 2. Participants must be able to vote 3. Participants must not have voted already 4. Participants must not have the intention to vote by mail 5. Participants must not report to be “extremely liberal” or “extremely conservative” In addition, we target only participants from certain US counties. This is because we cannot provide meaningful information about wait times for all counties. We will recruit participants from 414 counties which cover a total of 9,942 zip code areas. Participants who do not fulfill the requirements above or participants who do not come from the targeted counties are excluded from the study.

See above.

According to a power calculation, we need approx. 1,000 observations per treatment to detect a small effect in our belief measures (i.e., Cohen’s d = 0.12) with a power of 0.80 and an alpha of 0.05 (two-sided).