Voting Between Social Justice and Rewarding Pro-environmental Behaviour

The objective of this economic laboratory experiment is to assess the acceptability of environmental taxes on commuter mobility.
We investigate the impact of tax revenue recycling on tax acceptance. In an initial effort task, participants are divided into two groups with different travel costs.
the residents of these groups are asked to vote in favour of environmental taxes that increase the cost of travelling by private vehicle. This encourages people to reduce their use of these vehicles in favour of non-polluting public transport. To increase tax acceptance, we implement a number of treatments consisting of implementing different policies before the vote. These policies comprise a couple of tax and revenue recycling options.
We implement three different recycling options:
- Uniform redistribution: every participant gains an equal share of the tax revenue.
- Social redistribution: the tax revenue is used to compensate for the difference in travel costs between the two groups. If this difference is fully compensated for, the remaining revenue is redistributed equally among all participants.
- Environmental redistribution: the tax revenue is shared equally among public transport users.

We will test three main hypothesis:

Hypothesis 0 - Policy efficiency:
a. The private vehicle usage is lower when a tax is implemented, than when none are.

Hypothesis 1 - Tax implementation effect:
a. The acceptance of the lump-sum mechanism will increase after the first redistribution mechanism implementation.
b. The acceptance of the social cushioning mechanism will increase after the first redistribution mechanism implementation.
c. The acceptance of the environmental redistribution mechanism will increase after the first redistribution mechanism implementation.

Hypothesis 2 - Profit maximization:
a. Participant in population p will have higher acceptance for the cushioning mechanism.
b. Participant in population r will have lower acceptance for the cushioning mechanism.
c. The lump-sum and environmental redistribution treatments will have same level acceptance.

Our experiment also includes an assessment of various psycho-social constructs and inequity aversion coefficients.
We use these secondary outcomes to test the following hypothesis:

Hypothesis 3 - Social cushioning:
a. Participant in population r with high disadvantageous inequality (αi), will have lower acceptance for social cushioning.
b. Participant in population r with high advantageous inequality (βi), will have higher acceptance for social cushioning.
c. Participant in population p with high disadvantageous inequality (αi), will have higher acceptance for social cushioning.

Hypothesis 4 - Effect on bidding for taxes:
a. Participants with high levels of perceived legitimacy of redistribution mechanisms and environmental concern show greater support for environmental taxes, regardless of the redistribution mechanism.

Hypothesis 5 - Specific effect: :
a. Participants with low levels of system justification and subjective social status show a higher level of bidding for social redistribution.
b. Participants with high levels of environmental concern show greater support for environmental redistribution.

2025-09-23

2025-10-15

The experiment has two primary outcomes:
• The proportion of participants who chose private vehicles for their commute travel.
• Participants’ bids for the implementation of taxes.

- Inequity Aversion at the individual level
- Psychosocial constructs evaluation (MacArthur Scale of Subjective Social Status, Environmental concern, Inequality concern, Economic system justification scale, perceived legitimacy of redistribution mechanisms),
- Socio-demographic variables

20 sessions are organized, each with 20 participants. The participants are divided into 2 groups of 10.
Groups of ten participate together in an encryption game, competing to encrypt as many character strings as possible in five minutes. The five best participants are placed in a neighborhood with lower transport costs, while the others are placed in a neighborhood with higher costs.
Participants then take part in the main game.
In each round of the game, the participants are asked to choose between using a private car or public transport. The two options have different fixed costs. Using a private vehicle leads to congestion costs, which increase with the number of other vehicles in the same neighborhood, and less so with the number of vehicles in other neighborhood. he number of private vehicles affects all residents of both neighborhood through pollution cost. The Pigouvian tax used to reach the optimum has been calculated beforehand. We will implement different scenarios, each of which will correspond to a tax and a redistribution mechanism. These scenarios will represent our different treatments. Two sequences of scenarios (within subject design) will be implemented. In each sequence, the first scenario, without tax, will be used as a control. As this control scenario is common to both within-subject sequences, it act as a benchmark for comparison between sequences.

Experimental sequence A:
1. Trials (to account for learning effect, not used in the analysis):
• 2 rounds without taxes
2. Paid experiment:
• 5 rounds without taxes (control treatment)
• Votes on whether to implement a tax with uniform redistribution/ a tax with social redistribution
• 5 rounds with a tax with uniform redistribution
• Votes on whether to implement a tax with uniform redistribution/ a tax with social redistribution
• 5 rounds with a tax with social redistribution
• Votes on whether to implement a tax with uniform redistribution/ a tax with social redistribution
• 5 rounds tax and redistribution depending on a randomly selected votes outcomes.

Experimental sequence B:
1. Trials (to account for learning effect, not used in the analysis):
• 2 rounds without taxes
2. Paid experiment:
• 5 rounds without taxes (control treatment)
• Votes on whether to implement a tax with uniform redistribution/ a tax with environmental redistribution
• 5 rounds with a tax with uniform redistribution
• Votes on whether to implement a tax with uniform redistribution/ a tax with environmental redistribution
• 5 rounds with a tax with environmental redistribution
• Votes on whether to implement a tax with uniform redistribution/ a tax with environmental redistribution
• 5 rounds tax and redistribution depending on a randomly selected votes outcomes.

We implement a simulator on Otree allowing participants to play with taxes and the number of car users to see the effect on their costs. To test for a possible order effect, half of the groups (200 participants) are run with an inversion of the tax with uniform redistribution and the tax with the other redistribution (depending on the experimental sequence). Assignment to one or the other order is random. We obtain a 2X2 design : two different sequences and two different designs.

Voting system:
The voting system employs a bid mechanism analogous to that described by Messer et al. (2010):
1. Each (N) participant is endowed with a voting endowment (E), and is required to offer an amount (bi) to support a policy (taxes).
2. A random number R is generated between bmin and bmax. If the median bid is higher than R, then all participants pay R (won E-R), and the tax is adopted.
If the random number is lower than the median bid, the tax and the redistribution system are rejected, and all participants receive their inital endowment E.
Payments:
The remuneration is divided down into:
1. The remaining part of the voting endowment (two randomly drawn votes).
2. Two rounds randomly drawn among the 15 first ones an two among the last five rounds (where the vote is applied) are used to compute the gain, by subtracting the cumulative costs of congestion, and pollution from an initial endowment. The various taxes levied are deducted if the participant chooses to use a private vehicle, and then redistributed using the corresponding system.
3. Inequity aversion is remunerated.
4. All participants earn a flat participation fee.

More information can be found in the analysis plan.

Not available

The order and treatment assignment is randomized through computer using Python.

We consider two levels of randomization:
1) At the sequence level (i.e. experimental sequence A or experimental sequence B).
2) The order of the treatments within the sequence, excluding the control treatment ( i.e. whether or not the treatment with uniform redistribution comes before or after the treatment with the other redistribution mechanism).

Both the sequence and the order are randomized at the session level (two groups of 10 participants).
The same number of sessions will be held for each sequence and order (the total number of sessions will be divided into four equal parts).

Yes

A power analysis was conducted based on a forthcoming experiment (pre-registration available on the AEA RCT Registry website: https://www.socialscienceregistry.org/trials/13408). The analysis revealed that 40 groups of ten participants would provide sufficient power.


Further details can be found in the analysis plan.

There will be a total of 400 participants. There will be 200 participants for each sequence. For each sequence and each of the three scenarios (control, tax and uniform redistribution, and tax and alternative redistribution), participants will have five mobility choices and one vote for each redistribution system. Therefore : For experimental sequence A: • 5 rounds without taxes (control) : 5 * 200 = 1000 transportation choices • Votes on whether to implement a tax with uniform redistribution/ a tax with social redistribution : 200 votes for both redistribution mechanism • 5 rounds with a tax with uniform redistribution : 5 * 200 = 1000 transportation choices • Votes on whether to implement a tax with uniform redistribution/ a tax with social redistribution : 200 votes for both redistribution mechanism • 5 rounds with a tax with social redistribution : 5 * 200 = 1000 transportation choices • Votes on whether to implement a tax with uniform redistribution/ a tax with social redistribution : 200 votes for both redistribution mechanism For experimental sequence B: • 5 rounds without taxes (control) : 5 * 200 = 1000 transportation choices • Votes on whether to implement a tax with uniform redistribution/ a tax with environmental redistribution: 200 votes for both redistribution mechanism • 5 rounds with a tax with uniform redistribution : 5 * 200 = 1000 transportation choices • Votes on whether to implement a tax with uniform redistribution/ a tax with environmental redistribution: 200 votes for both redistribution mechanism • 5 rounds with a tax with environmental redistribution : 5 * 200 = 1000 transportation choices • Votes on whether to implement a tax with uniform redistribution/ a tax with environmental redistribution : 200 votes for both redistribution mechanism In each sequence, the number of observations per order, is the previous number divided by two. Secondary outcomes are assessed at participant level (400 observations per variables).

20 groups for sequence A, 10 groups for each order.
20 groups for sequence B, 10 groups for each order.

Analysis plan