Pivotal or Popular: The effects of social information and feeling pivotal on civic actions

Charitable Giving Field Experiment Intervention

All of these treatments followed the same basic format. Subjects were alumni of the education non-profit. On date 1 alumni were sent a message asking them to donate to the non-profit, those in the Control treatment were given a simple ask, while those in the four remaining treatments were told they would be randomly assigned to a group of 10 alumni and that if at least threshold T alumni donated in their group, then a third party donor would give an additional B bonus dollars to the non-profit. For the non-profit with the larger sample size the bonus, B, was $50; while for the other non-profit it was equal to $500.

We then let a few weeks pass so that alumni would have a chance to donate, and on date 2 we sent messages to those who had yet-to-donate which were meant to make them feel that donating was either (1) more/less popular, and (2) that their potential donation was either more/less important to reaching the threshold of T donors needed to obtain matching funds.

Voting Field Experiment Intervention
The day of the election we sent messages to all potential voters via email. This experiment only took place on a single date (unlike our charitable giving field experiment). To manipulate feelings about the popularity and importance of a person's vote we reported historical turnout rates, rather than updating potential voters about the current rates of voting.

2016-01-01

2019-12-31

If a person donated and the amount. If a person voted.

NA

NA

NA

Charitable Giving Field Experiment Intervention
1. Control: On date 1 alumni were sent a simple ask, and on date 2 a reminder to donate.
2. LessImpLessPop (T1 G1): On date 1 alumni were told they will be in a group of 10 alumni, and that if at least T = 1 persons donate, then an extra B bonus dollars will be given to the non-profit by a third-party. Then on date 2 they are told there has been only one gift (G1) so far. This is meant to signal that donation is not very popular. Furthermore they are told that the threshold goal of one donation (T1) has already
been met, and as such any donation they make would not be important or pivotal to reaching the threshold.
3. MoreImpLessPop (T2 G2): To signal greater importance while keeping the action of giving relatively less popular, in the date 1 message subjects are told they are in a group of 10 alumni but they now need T = 2 donors to reach the goal. Then on date 2 they are told there has been one gift so far. This is meant to signal that donation is not very popular. However, it also states that at the time of writing the group was
one gift away from reaching the goal, and as such it may signal that the recipient’s potential donation may be very important to reaching the goal for matching funds.
4. LessImpMorePop (T2 G2): Subjects in this treatment are sent the same letter on date 1 as those in the “MoreImpLessPop (T2 G2)” treatment. Then on date 2 they are told that two gifts have been received, making it impossible for their potential donation to
be important to reaching the goal while also showing that the act of donating is quite popular.
5. MoreImpMorePop (T3 G2): In our last treatment on date 1 subjects are told they need T = 3 donors to reach the goal, and on date 2 they are told that two donations have been received thus far. Thus, donation is a popular action, and the group is currently one gift away from reaching the goal making them feel important to reaching the goal for matching funds.
In our experiment we view a rate of two-out-of-ten gifts as signalling more popularity, and a rate of one-out-of-ten gifts as signalling less popularity. We communicate that a person is less important to helping the group obtain matching funds when the goal is already met
(G = T ), and signal more importance when the group is currently one person away from the goal (G = T −1).

Voting Field Experiment Design
1. Control: On the day of the election potential voters were sent a simple reminder to vote.
2. LessImpLessPop (T1 V1): Potential voters were informed that in 2016 the rate of voting had been one-in-ten (V = 1). Note, this was the true rate of voting in that year and is meant to signal that voting is not a very popular action. They were also told that if at least T = 1 persons voted in a randomly assigned group of 10, then their whole group would be entered in a raffle for a prize of B = 50 bonus dollars
provided by a third-party. Setting both the goal at one-in-ten and the previous rate at one-in-ten is meant to signal that this person’s vote is not going to be very important to reaching the goal. This message was chosen for the feelings it might evoke rather than the actual true underlying probability a person might infer from a base rate of one-in-ten as it is difficult to know the model that our potential voters are using in the background.
3. MoreImpLessPop (T2 V1): To signal greater importance while keeping the action of voting relatively less popular we informed voters about the same historical voting rate of one-in-ten (V = 1), but changed the threshold for being entered in the raffle to
two-in-ten (T = 2). We felt this might make subjects feel important to reaching the goal if they believe that their group will likely have exactly one voter and they need two voters.
4. LessImpMorePop (T3 V3): Potential voters were informed that in 2017 the true rate of voting had been three-in-ten (V = 3). Then to make them feel less important to reaching the goal to be entered in the raffle, we set the goal at three-in-ten members of their group voting. Much like the LessImpLessPop treatment the idea here is that subjects will use a heuristic that if the historical voting rate is the same as the goal,
they won’t personally need to vote for the goal to be met.
5. MoreImpMorePop (T4 V3): In our last treatment subjects face the same high rate of previous voting of three-in-ten, but the threshold for being entered in the drawing is raised to four-in-ten. We believe subjects will think that the group will have three other voters, making their vote very important to reaching the goal of four voters.

Randomization by computer

Individual

No

34,000 people

34,000 people

Control 6300, LessImpLessPop 4527, MoreImpLessPop 9605, LessImpMorePop 4517, MoreImpMorePop 9565