Rule Following and Cooperation

2022-07-26

2022-07-27

The number of tokens placed in the blue bucket are the primary outcome variable. They indicate the extent of rule following and cooperation, respectively.

** Study 2 (planned for March 2024)**

In the Normative Expectations sub-task, the primary outcome of interest is participants guess of the distribution of other participants personal norms with respect to the number of tokens placed in the blue bucket (i.e., participants guess of the distribution of other participants response to a question asking them what the most appropriate number of tokens is to place in the blue bucket).

In the Empirical Expectations sub-task, the primary outcome of interest is participants guess of the distribution of behaviour in previous waves (i.e., participants guess of the distribution of tokens placed in the blue bucket by previous participants).

In the Conditional Preferences sub-tasks, the primary outcome of interest is the conditionality of participants allocations to the blue bucket (i.e., how do participants condition their behaviour on the behaviour of others in the Conditional Preferences (Empirical) sub-task and how do participants condition their behaviour on the normative beliefs of others in the Conditional Preferences (Normative) sub-task.

In the Conditional Preferences sub-tasks, we apply the ABC methodology (Fischbacher & Gächter, 2010; Fischbacher et al., 2012; Gächter et al., 2017, 2022; Gächter & Marino Fages, 2023; Isler et al., 2021; Weber et al., 2023) to explain cooperation and rule following as a function of attitudes and beliefs.

Fischbacher, U., & Gächter, S. (2010). Social preferences, beliefs, and the dynamics of free riding in public goods experiments. American economic review, 100(1), 541-556.
Fischbacher, U., Gächter, S., & Quercia, S. (2012). The behavioral validity of the strategy method in public good experiments. Journal of Economic Psychology, 33(4), 897-913.
Gächter, S., Kölle, F., & Quercia, S. (2017). Reciprocity and the tragedies of maintaining and providing the commons. Nature human behaviour, 1(9), 650-656.
Gächter, S., Kölle, F., & Quercia, S. (2022). Preferences and perceptions in Provision and Maintenance public goods. Games and Economic Behavior, 135, 338-355.
Gächter, S., & Marino Fages, D. (2023). Using the Strategy Method and Elicited Beliefs to Explain Group Size and MPCR Effects in Public Good Experiments.
Isler, O., Gächter, S., Maule, A. J., & Starmer, C. (2021). Contextualised strong reciprocity explains selfless cooperation despite selfish intuitions and weak social heuristics. Scientific reports, 11(1), 13868.
Weber, T. O., Schulz, J. F., Beranek, B., Lambarraa-Lehnhardt, F., & Gächter, S. (2023). The behavioral mechanisms of voluntary cooperation across culturally diverse societies: Evidence from the US, the UK, Morocco, and Turkey. Journal of Economic Behavior & Organization, 215, 134-152.

Empirical expectations, that is, participants guesses of (1) the number of tokens other participants placed in the blue bucket and (2) the number of other participants who placed all of their tokens into the blue bucket.

We expect empirical expectations to be correlated with actual own decisions.

This experiment focuses on the relationship between rule-following and cooperative behaviour in social dilemmas, by adapting the “bucket task” by Kimbrough and Vostroknutov (2018). Across five experimental treatments participants are asked to allocate five tokens between a blue and a yellow bucket. In all experimental treatments the primary outcome of interest is the number of tokens placed in the blue bucket, which represents the rate of rule-following and cooperation, respectively. We will also elicit empirical expectations measured as the expected allocation of tokens to the blue bucket by other participants.

The five experimental treatments are:

1. Baseline Public Goods Game (PGG) – No Rule;
2. Baseline Rule Following Task (RFT) – No Public Good;
3. PGG with a Rule;
4. PGG with a Social Rule (“helps other group members”);
5. RFT and PGG No Rule (order of tasks randomised).

T1 to T4 are between-subjects, and T5 is within-subjects. All treatments are followed by a questionnaire asking about gender, age, and political ideology and the Cognitive Reflection Test (CRT; Frederick, 2005).

We will collect 104 participants in each of the 5 treatments, that is, a total of 520 participants. The experiment will be run on the platform Prolific using the software LIONESS Lab (Giamattei et al. 2020). It has received ethical approval from the Nottingham School of Economics Research Ethics committee.

Both the PGG and the RFT are one-shot games.

** Follow up Experiments of T5 (planned for December 2022) **

In the first wave of experiments with T1 to T5 (run on Prolific on 27 July 2022) we collected about 104 observations per treatment on average. In T5 which has two orders we only had 56 observations on average. Since there is evidence for order effects with the RFT we decided to replicate T5 with increased numbers of observations such that we have the same expected number of observations in all treatments regardless of order.

We will therefore rerun T5 (RFT and PGG No Rule) in both orders with an expected sample size of 110 each. This will give us increased power to detect any potential order effects, and also allow for proper comparison between behaviour in T5 with behaviour in Baseline PGG and Baseline RFT respectively. To be on the safe side the target sample size in both orders of T5 is 110. To allow for dropouts (estimated from wave 1 to be roughly 10%) and to facilitate the collection of equal numbers our target will be to recruit 250 participants (that is 125 per order in T5) which should accommodate for potential dropouts of about 15 per treatment enabling us to collect 110 complete observations in each order.

All procedural details described in the rest of this preregistration remain the same.

** Follow up Experiments of Baseline RFT (planned for June 2023) **

In a follow up study to the first wave of experiments with T1 to T5 (run on Prolific on 27 July 2022) we will collect 464 observations across 4 treatments, which maintain the structure of the initial experiments described above.

In this follow up study we will conduct 4 treatments which replicate and extend Baseline RFT in the first wave, implementing a 2x2 design.
We will examine how behaviour in the decision task in Baseline RFT is impacted by removing the stated rule and/or by the addition of a sentence in the instructions stating participants submissions will be accepted if they submit the correct completion code for the study.

We have two main hypotheses which will be tested by this additional experiment. Hypothesis 1: We expect no difference in behaviour due to the addition of a sentence in the instructions in treatments with a rule. Hypothesis 2: The proportion of tokens placed in the yellow bucket by participants should be significantly higher in treatments without a rule.


** Follow up Study 2 (planned for March 2024) **

In a follow up study to the first wave of experiments with T1 to T5 (run on Prolific on 27 July 2022) we will collect 3,080 observations across 20 experimental treatments. Each experimental treatment comprises one Task and 2 Sub-tasks.

The target sample size in all treatments is 140 participants. We recruit 3,080 participants in total to account for potential dropouts (estimated at roughly 10% based on previous waves) and facilitate equal numbers across treatments. We exclude any incomplete data, and for all reported analysis we consider only the first 140 complete submissions in each treatment from participants who submitted a completion code on Prolific. Therefore, the total sample considered in reported analysis will be 2,800 participants (140 participants x 20 treatments).

In this follow up study we extend the structure of the first wave, to examine social expectations and conditional preferences. We will examine how normative expectations, empirical expectations, and conditional preferences (both normative and empirical) differ across treatments with and without a rule/public good. We consider 5 tasks which mirror the tasks described in previous waves:
- T1.Baseline RFT;
- T2. Baseline RFT No Rule;
- T3. Baseline PGG;
- T4. PGG with Rule;
- T5. PGG with Social Rule.

The follow up experiments are between-subjects and participants are randomly assigned to one of these tasks. Participants complete 2 sub-tasks during the experiment. For all participants, the Norm-Drawing Task (Panizza, F., Dimant, E., Kimbrough, E. O., & Vostroknutov, A. (2023). Measuring Norm Pluralism and Tolerance. Available at SSRN.) constitutes one of their sub-tasks.

For their other sub-task, participants are randomly allocated to one of the following four sub-tasks: ST1. Normative Expectations, ST2. Empirical Expectations, ST3. Conditional Preferences (Normative), ST4. Conditional Preferences (Empirical).

The purpose of this follow up experiment is to measure social expectations and conditional preferences in the respective (sub-)tasks and analyse how they are influenced by the presence and nature of a rule. We hypothesise that rule following and cooperation should exhibit different contribution schedules for conditional preferences, the presence of a rule should increase normative expectations about the most appropriate allocation to the blue bucket and increase empirical expectations of previous participants allocations to the blue bucket.

Both the PGG and the RFT are based on the “bucket task” by Kimbrough, E. O., & Vostroknutov, A. (2018). A portable method of eliciting respect for social norms. Economics Letters, 168, 147-150. In our version of this task, people have five tokens and are asked to allocate the five tokens to either a “blue bucket” or a “yellow bucket”. In the RFT tokens placed in the blue bucket are worth 0.5 points and tokens placed in the yellow bucket are worth 1 point. Participants are paid according to their decisions, where 1 point = £0.20. Economic self-interest therefore suggests that people place all tokens in the yellow bucket. The RFT simply stipulates our rule: “The rule is to place the tokens in the blue bucket”. Following this rule is therefore a costly act and violating it in people’s self-interest.

The number of tokens placed in the blue bucket is a measure of the extent of rule following: 5 tokens in the blue bucket implies complete rule following; 0 tokens in the blue bucket signifies complete rule violation.

The PGG is the same as the RFT except that the PGG is played in groups of four participants and the RFT is an individual task. In the PGG group members will receive a payoff from tokens contributed by all group members. The sum of tokens allocated by all group members to the blue bucket are doubled and each group member receives 1 quarter of the value of the tokens, regardless of their own contributions. Participants completing the PGG task only see a rule if they are in either the PGG with a Rule or PGG with a Social Rule treatments. In PGG with a Rule the rule is the same as in the RFT above ("The rule is to place the tokens in the blue bucket"), and in PGG with a Social Rule the rule stipulates: "The rule is to place the tokens in the blue bucket because it helps the other group members".

After completing their task participants provide their empirical expectations of others behaviour, by providing their guesses of (i) the number of tokens placed in the blue bucket by other participants; and (ii) the number of other participants who placed all of their 5 tokens in the blue bucket. In RFT and PGG participants complete both tasks, with the order randomised, and provide their empirical expectations after each task. Participants empirical expectations are then compared to their actual behaviour during the task(s).


** Follow up Study 2 (planned for March 2024) **

This experiment is between-subjects. Participants complete only 1 task. Possible tasks are: (1) Baseline RFT; (2) Baseline RFT No Rule; (3) Baseline PGG; (4) PGG with Rule; (5) PGG with Social Rule. Within their given task participants complete 2 sub-tasks. For all participants, Norm-Drawing (Panizza et al., 2023) constitutes 1 of their 2 sub-tasks. For the other sub-task participants complete only 1 sub-task from the following: (i) Normative Expectations; (ii) Empirical Expectations; (iii) Conditional Preferences (Normative); (iv) Conditional Preferences (Empirical).

Guaranteed Payment(s)
- Participation Fee: £2.50

- Conditional Preferences (Normative) sub-task:
Contribution: For 1 (out of 4) group members conditional choice chosen, other 3 group members unconditional choice chosen, paid as in
Study 1 for that choice (yellow token=1 point=£0.20, blue token=0.5 points=£0.10) so those assigned to (1) Baseline RFT or (2) Baseline
RFT No Rule guaranteed between £0.50-£1.00, while those assigned to (3) Baseline PGG, (4) PGG with Rule or (5) PGG with Social
Rule guaranteed between £0.50-£2.50

- Conditional Preferences (Empirical) sub-task:
Contribution: For 1 (out of 4) group members conditional choice chosen, other 3 group members unconditional choice chosen, paid as in
Study 1 for that choice (yellow token=1 point=£0.20, blue token=0.5 points=£0.10) so those assigned to (1) Baseline RFT or (2) Baseline
RFT No Rule guaranteed between £0.50-£1.00, while those assigned to (3) Baseline PGG, (4) PGG with Rule or (5) PGG with Social
Rule guaranteed between £0.50-£2.50

Potential Bonus Payment(s)
- Norm-Drawing sub-task:
£1.00 if guess sufficiently close to average distribution made by other participants, £0.00 otherwise

- Normative Estimation sub-task:
£2.00 if guess of other participants (out of 100) sufficiently close to true distribution of personal norms, £0.00 otherwise

- Empirical Estimation sub-task:
£2.00 if guess of other participants (out of 100) sufficiently close to true distribution of behaviour, £0.00 otherwise

- Conditional Preferences (Normative) sub-task:
(Beliefs: £2.00 if guess of other group members (out of 3) personal norms sufficiently close to true distribution, £0.00 otherwise)

- Conditional Preferences (Empirical) sub-task:
(Beliefs: £2.00 if guess of other group members (out of 3) behaviour is sufficiently close to true distribution, £0.00 otherwise)

People will be randomly allocated into one of the five treatments by the LIONESS Lab software (https://lioness-lab.org/).

In the follow up experiments of T5 participants will be randomly allocated into one of the two orders by the LIONESS Lab software (https://lioness-lab.org/).

In the follow up experiments of Baseline RFT participants will be randomly allocated into one of the four treatments by the LIONESS Lab software (https://lioness-lab.org/).

In the follow up experiments of Study 2 participants will be randomly allocated into one of the 20 treatments by the LIONESS Lab software (https://lioness-lab.org/).

Individuals are randomly allocated to one of the 5 treatments. In T5 the order of tasks is randomized, and the task that is paid will be randomized as well.

In the follow up Study 2 individuals are randomly assigned to one of the 20 treatments.

Yes

In treatments 1 - 4 participants will take only one decision. In T5 participants take two decisions and we will therefore cluster at the individual level.

We will recruit 104 participants for each of the 5 treatments, that is, a total of 520 people. The previous sentence refers to the previous experiment run in July 2022. We now plan a follow up with an additional 250 participants, allowing for dropouts. See details above in the experimental design. The planned follow up experiment in June 2023 will recruit 464 participants. ** Follow up Study 2 (planned for March 2024) ** We will collect 3,080 observations across 20 experimental treatments. The target sample size in all treatments is 140 participants. We recruit 3,080 participants in total to account for potential dropouts (estimated at roughly 10% based on previous waves) and facilitate equal numbers across treatments. We exclude any incomplete data, and for all reported analysis we consider only the first 140 complete submissions in each treatment from participants who submitted a completion code on Prolific. Therefore, the total sample considered in reported analysis will be 2,800 participants (140 participants x 20 treatments).

On average 104 participants per treatment were collected in July 2022.

In the follow up experiments of T5 we plan to recruit 125 participants with a target number per order of 110 after dropouts.

In the follow up experiments of Baseline RFT we plan to recruit approximately 116 participants per treatment.

** Follow up Study 2 (planned for March 2024) **
The target sample size per experimental treatment is 140 online Prolific participants.