Expert Beliefs and Community Organizers' Demands

2023-08-03

2023-08-19

Willingness to Volunteer; Willingness to Volunteer-Support; Willingness to Volunteer-Oppose; Willingness to Sign

(1) Willingness to Volunteer: a binary variable that takes the value one if the respondent is
willing to volunteer (support or oppose the technology), zero otherwise.; (2) Willingness to Volunteer-Support: a binary variable that takes the value one if the respondent is willing to volunteer to support the technology, zero otherwise. (3) Willingness to Volunteer-Oppose: a binary variable that takes the value one if the respondent is willing to volunteer to oppose the technology, zero otherwise; (4) Willingness to Sign: a binary variable that takes the value one if the respondent is willing to sign a petition to oppose the technology, zero otherwise

Quality of Volunteering Task

Quality of Volunteering Task: We plan to conduct an open-text analysis of the free response section of the survey. We will use this analysis to measure the quality of responses across various
participants and code their strength of support or opposition to the technology. If participants
choose either Old Firm Name or New Firm Name, we ask them to fill out a free-form text box
with reasons as to why they support that specific cause. We mention that this text will be sent
to advocates in either cause to use to build their database. Research assistants will independently
classify open-text responses (Duke undergraduate students). If there is a disagreement between
coders, the classification selected by the majority will be used.

We will conduct the analyses in a few steps. Noting that we already have several baseline characteristics for the individuals who have previously made predictions, we will not directly measure any
covariates in our survey.

This survey will follow the following steps: first, participants will be reminded of the predictions
they made previously and what the question they were asked previously. We will then randomize
participants into one of the two treatment arms, where they will receive information regarding how
experts predicted the price change. We then introduce participants into Old Firm Name Treatment
or New Firm Name Treatment, give them a brief overview of the company and the cause, and ask
them if they would be willing to volunteer or sign the open letter.

Not available

We use the ‘evenly present elements’ functionality in Qualtrics.

individual level

Yes

Since we previously conducted this survey and received a total of 2,500 responses, this survey will
be bounded by that sample

No more than 2,500 responses

First, we randomize participants into two groups (50% in each treatment arm): (1) Control and
(5) All Information. To conduct randomization, we use the ‘evenly present elements’ functionality
in Qualtrics. The two treatment conditions are: (1) Control: receive information on the true price change and own prediction, (2) All Treatment: Receive information on the relative accuracy of all experts (finance, police
officers and community organizers) as well as the true price change, and own prediction.

Next, we randomize if people see Old Firm Name and Stop - Old Firm Name as the name of
the cause or the new name New Firm Name and Stop - New Firm Name: (1) Old Firm Name Treatment (50%): All of the text regarding the specifics of the company and the cause will follow the ShotSpotter/Stop ShotSpotter verbiage; (2) New Firm Name Treatment (50%): All of the text regarding the specifics of the company and the cause will follow the SoundThinking/ Stop SoundThinking verbiage.

We run power calculations using the pwr package in R. We do not have strong priors on what the size of effects will be. We the fix type II error (power) at 0.8, the type I error ($\alpha$) at 0.05, and compute the required sample size so that our primary analysis will have a be powered to detect a Cohen's D and Cohen's H of 0.2. (The sample size was based on a previous study)