Non-compliance and the voluntary provision of public goods: the role of cost uncertainty and heterogeneity

2021-11-08

2022-03-31

We investigate the treatment effects on pledged and actual contribution levels. Comparing these two, we additionally investigate compliance levels.
Hypothesis 1: Cost-heterogeneity induces high cost types to pledge lower amounts and reduces contribution and compliance levels.
Hypothesis 2: Cost-uncertainty reduces the pledges and leads to lower compliance and contributions for high-cost types.
In terms of the review stage that is added to the pledge mechanism in the standard public good game, we intend to determine the effect that reviews (i.e. mutual grading) have on initial pledges as well as compliance and contribution behavior. We hypothesize that a mutual grading process may reduce pledges under uncertainty, but overall induces higher compliance and contribution levels.
Hypothesis 3: A review process incentivizes more cooperative behavior, i.e. more compliance with the initial pledge and/or higher contribution levels.
We also expect of contributions to depend on both the pledges and the contributions costs: high costs types are graded more lenient when not adhering to their pledges than low cost types.
We additionally explore how pledges and contributions depend on whether they refer to a quantity or costs.
Hypothesis 4: In comparison to a contribution pledge, cost pledges lead to larger compliance levels under costs uncertainty and heterogeneity. They yield more equal payoffs among heterogeneous actors as the focus is rather placed on the equal shares contributed than on the unit quantities contributed.

First, we intend to compare the effects by using nonparametric tests. Data are aggregated at the group level. Between treatments, pledges, compliance, contributions are compared across all subjects as well as by subjects conditional on cost types (Mann Whitney U). Within treatments with costs heterogeneity , pledges, compliance, contributions are compared between cost types (Wilocoxon signed rank). Second, we perform difference-in-differences regression analyses. Third, we look at the heterogeneous treatment effects to see if there are interactions between control variables (see below) and treatment effects. The analysis of possible heterogeneous treatment effects will be explorative.

Control variables from a short questionnaire at the end of the survey:
• Demographic variables (age, gender, education level)
• Social preferences (risk taking, importance of own compliance, importance of compliance others),
• Task comprehensibility.

This experiment will be conducted as an online experiment of the experimental laboratory of the University of Hamburg. The experiment will make use of a linear public good game with a 2x4 (+1) between-subject design. More precisely, the subjects will be assigned to groups of four players who interact in a five-round public good game. A pledge process (as in Barrett and Dannenberg, 2016), which requires the players to make a pledge about their subsequent contribution, extends the standard public good game of this experiment.
The pledge-making phase is followed by five contribution rounds of a standard public good game.
In treatment 1, all subjects have identical costs, which are known when they make their pledge. Half the groups feature high, half low costs.
Treatment 2 implements heterogeneous costs, i.e. two players are high-cost types and two players are low-cost types, which they learn before making their pledge.
Treatment 3 features cost-uncertainty at the moment of the pledge-making. The homogeneous costs for the group – they either all have high or low costs per unit of contribution - are revealed after the players made their pledges, but before the five contribution rounds.
Treatment 4 combines the two effects of cost heterogeneity and cost-uncertainty. Thus, at the moment of pledge making, the players do not know who has high and who has low costs.
All these treatments are in a variant A (1A through 4A) and B (1B through 4B): In variants B, the pledge mechanism is extended to a pledge and review (P&R) mechanism. Firstly, players are reviewed for their pledges and secondly, for their contributions after every round. Every player reviews all players and is reviewed by the others in form of a school grade 1-6 (1: very good, 2: good, 3: satisfactory, 4: ausreichend, 5: ungenügend, 6: mangelhaft), as known from the German school system. After the review stage, the grade of a player is calculated as the average of the three grades given by the other players.
All treatments provide full transparency, i.e. all players know each other’s costs, pledges, contributions and in the case of the B treatments, their grades received.
One further explorative treatment (4C) is analogous to treatment 4B, cost-heterogeneity and cost-uncertainty under a P&R mechanism, with the difference that players do not make a contribution pledge regarding their contribution units but a cost pledge in which they state how much money they are willing to bear for the public good.
Finally, respondents are required to complete a short questionnaire to collect data on possible control variables. Questions include demographic variables (age, gender, education level) and social preferences questions (risk taking, trust, altruism, positive reciprocity, negative reciprocity).

Randomization done in office by a computer

Individual

No

-

In all treatments, subjects interact in groups of 4. Each group of 4 generates an independent observation. We once run the experiment with a smaller cost heterogeneity (0.7 vs. 1.3) and once with a bigger cost heterogeneity (0.6 vs. 1.8). For the first round of sessions with the smaller cost heterogeneity we plan for 12 groups (48 subjects) per treatment. As treatments 1A,1B,3A,3B are subgrouped into groups with high and low costs, we plan to generate the double amount, i.e. 24 groups (96 subjects). Generating a total of 624 subjects. These sessions are run in November and December 2021. For the second round of sessions with the bigger cost heterogeneity we plan for 15 groups (60 subjects) per treatment. Corresponding to before, we take the treatments 1A, 1B, 3A, 3B double, i.e. 30 groups (120 subjects), which generates a total of 780 required subjects. These sessions are run in February and March 2022.

-