Time-Based and Milestone-Based Feedback in Public Goods Provision (Online Laboratory Experiments)

2021-11-09

2023-12-31

1) Total group contributions at the end of each sequence
2) Total group contributions at the end of each round within the sequence
3) Proportion of groups with total contributions meeting specific thresholds/milestones at the end of each sequence

1) Individual contributions by group members in each round within the sequence

[Study has been withdrawn as we decided to implement the experiment face-to-face instead of online, given the easing of COVID-19 restrictions in Australia.]

Participants take part in five sequences of a public goods game with dynamic contributions. At the beginning of each sequence, participants are divided into groups of four (random rematching across sequences), and each group member is given an endowment of 50 tokens. Each sequence consists of 10 rounds where, in each round, group members can decide how much of their remaining endowment to contribute to the Group Account. Any unassigned tokens at the end of the 10th round will remain in the members' Private Accounts. Group members receive their payoffs from the Group and Private Accounts only at the end of the 10th round.

Payoffs follow the standard VCM experiment: For every token remaining in a group member's Private Account, that member receives 1 token. For every token there is in the Group Account, each group member receives 0.4 tokens.

The main treatments vary on the information that group members receive at the end of each round of the sequence, with feedback being either time-based or milestone-based. Where feedback is milestone-based, we will also vary the intervals of the milestones.

Participants take part in five sequences of a public goods game with dynamic contributions. At the beginning of each sequence, participants are divided into groups of four (random rematching across sequences), and each group member is given an endowment of 50 tokens. Each sequence consists of 10 rounds where, in each round, group members can decide how much of their remaining endowment to contribute to the Group Account. Any unassigned tokens at the end of the 10th round will remain in the members' Private Accounts. Group members receive their payoffs from the Group and Private Accounts only at the end of the 10th round.

Payoffs follow the standard VCM experiment: For every token remaining in a group member's Private Account, that member receives 1 token. For every token there is in the Group Account, each group member receives 0.4 tokens.

The main treatments vary on the information that group members receive at the end of each round of the sequence:
- In Treatment T (Time), participants are informed how many tokens have been accumulated in the Group Account at that point in the sequence.
- In Treatment M25 (Milestone-25), participants are informed whether the accumulated number of tokens in the Group Account has met the thresholds of 25, 50, 75, 100, 125, 150, 175, or 200.
- In Treatment M50 (Milestone-50), participants are informed whether the accumulated number of tokens in the Group Account has met the thresholds of 50, 100, 150, or 200.

After the main task is completed, participants also complete a conditional cooperation task (similar to Fischbacher et al., 2001 and Fischbacher and Gächter, 2010). Participants will also be asked to complete a questionnaire eliciting their demographic characteristics, reasons for their decisions in the experiment, as well as decisions in a one-shot risk task (Gneezy and Potters, 1997).

Randomization is done at the session level. Treatment assignment is randomly pre-determined before the session is run, and participants select into sessions randomly without knowing the pre-determined treatment assignment.

Unit of randomization is at the experimental session level

Yes

21 to 24 (depending on show-up rate for each session)

480

160 (between 7 and 8 sessions each with 20 or 24 participants, depending on show-up rate)

We consider effect sizes for group contributions at the end of each sequence. The main unit of observation in our analysis is at the group-sequence level. With each group of 4 participating in 5 sequences, 160 participants yield 40 groups or, equivalently, 200 group-sequences. For a Type I error rate of 0.05 and statistical power of 0.80 under a two-tailed test, this gives us an effect size of 0.287 standard deviations (between small and medium).