Do election campaigns make leaders more responsible?

We use an asymmetric between-subjects design where the treatments vary on two dimensions. First, the treatments vary whether candidates can campaign on their own contribution to the common fund or not. Second, the treatments vary whether their campaign messages are binding or not. Binding campaign messages are automatically implemented whereas candidates get a chance to revise their campaign message if they are non-binding. The experiment will be conducted with participants recruited using OSREE at the E2MU lab at the University of the Melbourne.

2025-03-18

2026-12-31

1- Tax Amount stated in the Campaign Message
2- Own Contribution stated in the Campaign Message (if applicable)
3- Actual Tax Amount
4- Actual Own Contribution

1- We will create a variable called "promise keeping" using primary outcomes 1, 2, 3 and 4.
2- Payoff to players calculated using primary outcomes 3 and 4 above.
3- A variable representing leader type will be created using the following categorization: Embezzlers (those who take money from the common fund); Tax evaders (those who contribute less than the tax amount); Responsible (those who contribute at least as much as the tax amount).

1. Elected leader's campaign choices
2. Elected leader's level of promise keeping

In our experiment, participants are randomly assigned to groups of five. Within each group, three participants are assigned the role of voters, while the remaining two are assigned the roles of Candidate. All participants receive an endowment. The candidates in each group send campaign messages to the voters. The campaign messages contain information about the tax amount. In one treatment, it also includes information about the leader’s contribution to the common fund. The campaign messages may be binding or non-binding, depending on the treatment condition. If the campaign message is binding, its contents are automatically implemented if the candidate is elected to be the leader. After the voters receive the campaign messages from the candidates, they vote for their preferred candidate. While voters are casting their votes, candidates make their final decisions about the tax amount and their own contribution to the common fund (in the case when the campaign messages are not binding). Hence, we use a strategy method and ask all candidates to state their decisions which are implemented if they are elected to be the leader. The leader is determined by majority voting. The elected leader and all group members receive their initial endowment plus a share of the total common fund whereas the non-elected candidate simply receive their initial endowment.


The participants also complete a survey, which consists of three parts. In Part I, the participants provide feedback on their understanding of the instructions and payoff calculations. They are also asked open-ended questions about the decisions they made during the experiment and their preferences (honesty and guilt aversion). In Part II, participants are randomly re-grouped into groups of five for a decision-making task. Each participant is asked to split 100 ECU between themselves and the other group members, with the allocated amount being equally distributed among the other four participants. At the end of the session, one participant from each group is randomly selected, and their decision is implemented. In Part III, the participants answer demographic questions, including gender, age, field of study, years in Australia, ethnicity, and past participation in experiments.

In our experiment, we classify three types of leader behaviors. First, if the leader makes a negative contribution to the common fund (i.e., takes form the common fund, the behavior will be classified as embezzlement. Second, if the leader makes a lower positive contribution to the common fund than the tax rate imposed on voters, the behavior will be classified as tax evasion. Finally, if the leader makes a positive contribution that is equal to the tax rate or exceeds it, the behavior will be classified as responsible.

Our main research questions are:
1. Does making campaign promises binding make elected leaders more responsible?
2. What is the impact of making campaign promises binding on voter welfare and total welfare?
3. Do non-binding campaign promises make elected leaders more responsible?
4. What is the impact of non-binding campaign promises on voter welfare and total welfare?

Not available

Participants will be recruited via email using ORSEE. They will choose between a list of available sessions and the session is randomized to a treatment before it is initialized.

Roles are randomly assigned by the computer within each session. Depending on the randomized treatment, candidates can either campaign on tax rate alone or both tax rate and their own contribution to the common fund. Similarly, the treatment will determine whether their campaign messages are binding or not.

Treatments are assigned at the session level whereas the roles are randomized at the individual level within a group.

No

450 Participants.

Based on the power calculations, we need 450 (150 x 3 Treatments) participants in total.

We plan to run three treatments. For each treatment arm, we plan to run 6 sessions (5 groups or 25 participants within each session). Hence, we plan to recruit 150 participants (or randomly assigned 60 candidates and 90 voters) per treatment arm.

The power calculations for each of the following main outcome variables based on the selected test are as follows: (1) Tax Amount Campaigned: Variable Type: Ordinal, Test: Wilcoxon test for mean comparison for two-independent groups, Level of Significance: 5 Percent and Power: 80 Percent, Sample Size: 61 candidates per treatment (~30 groups), Effect Size (d): 0.46, Minimum Detectable Effect Size: 7 ECU. (2) Leader's Own Contribution: Variable Type: Discrete, Test: Wilcoxon test for mean comparison for two-independent groups, Level of Significance: 5 Percent and Power: 80 Percent, Sample Size: 53 candidates per treatment (~27 groups), Effect Size (d): 0.5, Minimum Detectable Effect Size: 25 ECU. (3) Voter's Payoff: Variable Type: Continuous, Test: Wilcoxon test for mean comparison for two-independent groups , Level of Significance: 5 Percent and Power: 80 Percent, Sample Size: 53 candidates per treatment (~27 groups), Effect Size (d): 0.5, Minimum Detectable Effect Size: 12 ECU. (4) Leader Type - Responsible: Variable Type: Dummy, Test: Fisher's Exact test for inequality in proportion for two-independent groups, Level of Significance: 5 Percent and Power: 80 Percent, Sample Size: 61 candidates per treatment (~30 groups), Minimum Detectable Effect Size: 0.20.