Who Gives a Dime? Experimental evidence on first-movers in fundraising mechanisms.

We vary the use of prize mechanisms to increase individual contributions to a public goods game in the simultaneous and sequential setting. The control treatment is the voluntary contribution mechanism (VCM) with no prize. The treatments are a pay-your-bid lottery (LOT) and all-pay auction (APA) with an equally sized prize.

Participants play a public goods game in groups of three over fifteen rounds. As our treatment we vary the prize mechanism. In the control condition (VCM) there is no prize, in the lottery treatment (LOT) each unit of contribution buys a lottery ticket and the prize is awarded by raffle, in the all-pay auction treatment (APA) the prize goes to the highest contributor. Each session uses one treatment throughout. Within each round participants play both a simultaneous and a sequential version of the game, allowing direct comparison of contribution behavior across decision protocols under each prize mechanism.

2026-05-06

2026-05-07

Individual average contributions, Public good provision (gross and net of prize)

Expected average contributions of group members (first-order beliefs on others' contributions), Voting outcome of final round.

The experiment combines a between-subjects design across prize mechanisms (VCM, LOT, APA), assigned at the session level, with a within-subject design across timing structures, as every participant plays both the simultaneous (SIM) and sequential (SEQ) protocol within each round. Groups of three are randomly rematched every round. The sequential protocol uses the strategy method, eliciting a complete conditional response schedule from every participant before roles are assigned. Incentivized beliefs are collected after each decision. Payment is determined by a scheme in which one decision round and one prediction round are drawn at the end of the session, applying equally to all participants and preventing hedging. A final incentive-compatible majority vote determines a bonus round. With one session per treatment the study is exploratory in nature.

Our experiment combines two design dimensions. The prize mechanism (VCM, LOT, APA) is manipulated between subjects at the session level, as each session is assigned to one treatment and all participants face the same prize mechanism throughout. The timing structure, simultaneous (SIM) versus sequential (SEQ), is manipulated within subjects. Every participant plays both protocols in every round, with the order counterbalanced across groups within each session. Half the groups play the simultaneous decision first followed by the sequential decision, while the other half play the sequential decision first to minimize order effects. This within-subject variation across timing structures allows direct individual-level identification of the effect of observability and leadership on contributions, controlling for individual heterogeneity, group composition, and treatment. The between-subjects treatment assignment implies the session is the unit of randomization and constitutes a single cluster per treatment arm. The study is therefore best characterized as exploratory with limited statistical power for between-treatment comparisons.

Each of the 15 rounds consists of two parts. In one part, all participants decide simultaneously and independently on what to contribute to the public good, no participant observes another's decision before submitting their own. In the other part, the strategy method is used. Here, every participant submits both a first-mover contribution and a complete conditional response schedule specifying their contribution as second mover for each possible first-mover transfer. Roles are then assigned randomly within each group, where one player is designated the first mover and the other two are second movers, and realized contributions are determined by the first mover's decision and the second movers' schedules evaluated at that decision. Within each session participants are randomly rematched into groups of three every round, minimizing repeated-game incentives and collusion. After each decision task participants submit incentivized predictions about the average contribution of the other two group members. In Part 2 they additionally predict the others' full second-mover schedule. Each correct prediction earns a fixed bonus. A random-round incentive scheme ensures that every round is decision-relevant. At the end of the 15 rounds one round is drawn at random to pay decision earnings and a separate round is drawn to pay prediction earnings, with the same rounds applying to all participants in the session. An additional bonus round is played using the mechanism selected by majority vote within each group, and its decision earnings are added to the final payment. The vote is incentive compatible since the bonus round is paid out at the same rate as regular rounds. Participants are recruited from the University of Amsterdam (UvA) subject pool and sessions are conducted in a controlled laboratory environment at UvA's CREED.

Subjects are randomized into one of three sessions. In the laboratory, randomization in each session occurs by computer (in the oTree software) as soon as participants begin the experiment on the screen. Each round participants are rematched to other players, with the possibility of playing against the same participant, again by computer.

Experimental session for between-subject analysis across treatments, Individual for within-subject analysis across timing structures.

Yes

3 sessions.

1215 observations (81 individuals x 15 rounds).

81. From 27 participants in the VCM, 27 in the LOT, and 27 in the APA treatment.

MDE is approximately 1.3 tokens (26% of endowment) with an assumed standard deviation of 1.75 tokens.