I think that authors sufficiently addressed my concerns and the registered report is ready for data collection.

I thank the authors for the detailed response to my previous comments. According to my assessment the revised manuscript indeed represents a much better aligned research design. I believe all of my previous concerns have been sufficiently addressed in the revision. In the following, I would only like to suggest some minor changes to the experimental instructions.

To clarify what happens each round and what is based on all decisions, I would suggest the following general changes:

“2. Company’s Income and Income Taxation” Change this to “2. Company’s Income and Income Taxation each round”

“3. Company Success” Change this to “3. Company Success each round”

“3. Audit of the Reported Income” Change this to “3. Audit of the Reported Income each round”

Alternatively, you could also enter an introductory sentence to clarify that income and success calculations happen on a round-by-round basis just to be sure that participants do not mix this up as they only receive certain information in the end.  

I further have some minor wording suggestions in the specific treatment instructions to enhance clarity:

2.2 Instructions Gr-No & Gr-Ind & Ind–Gr & Gr-Gr

“You will only receive information on the other employee's report at the end of the experiment.” Change this to: “You will only receive information on the other employees’ reports at the end of the experiment”

“The tax will be calculated based on the  average report of the income that you  and your two colleagues report” Change this to: “The tax will be calculated based on the  average of the incomes that you  and your two colleagues report”

2.3 Instructions Ind-Ind

“The company success is 1000 for a tax of 0 (reported income 0). If  audited  the unpaid  tax needs to be paid (250) and a fine of the same amount (250) resulting in a company  success of 500 (1000  -250  -250).” Change this to “The company success is 1000 for a tax of 0 (reported income 0) if your report is not audited. If  audited  the unpaid  tax needs to be paid (250) and a fine of the same amount (250) resulting in a company  success of 500 (1000  -250  -250).”

2.4 Instructions Gr – Ind & Gr-Gr

Could it be that in each of these instructions the example provided under “4. Company success” is missing information on the audit?

​

The submitted registered report addresses a relevant and novel research question. Personally, I would be very interested to see the results of the suggested experiment. Furthermore, I believe that tax compliance is a very good practical example to assess the impact of individual versus group punishment. However, I struggle with the theoretical placement of the research question and the theoretical decision model. Therefore, I have a couple of comments and questions regarding the theoretical argumentation and their fit to the experimental design. Probably most of my questions are the result of my personal shortcomings in understanding the provided material and I apologize for that. Nevertheless, the points raised need clarification before I can issue my final assessment. Overall, I hope the authors will find my feedback helpful to further develop their research.

1)        To regard tax compliance as a mere question of dishonesty does not fully capture the problem. Fochmann et al. (2021) explain to some length that tax compliance combines dishonesty with risk-taking. I believe the submitted report would benefit from taking this more specific description of the tax compliance decision into consideration. There is a large literature on risk-taking by groups that currently is barely considered by the authors. See for example the work by Gary Charness and co-authors.

2)        Many experimental studies have investigated the effect of punishment, i.e., different audit and penalty rates, on individual tax compliance. None of these studies is currently mentioned in the literature overview on the effects of punishment. This literature seems to be very relevant to the suggested study. It provides ample causal evidence on the effects of punishment and it is perhaps more closely related to the experimental design than the literature on punishment in general that is currently considered by the authors. For an overview of relevant experimental studies see Alm and Malézieux (2021).

3)        In my opinion, the first contribution is overstated. This contribution has already been made by Bonfim and Silva (2019) as well as Fochmann et al (2021). Regarding Fochmann et al. (2021): It is true that punishment is not manipulated in the underlying experiment. However, the authors do manipulate the group setting. Therefore, the effect of the group context on decision making can be assessed exactly. The effect of the risk of punishment can be assessed in comparison to prior studies on dishonesty that did not include risk. This is the contribution the authors make. The current paper verifies this finding by integrating a “no risk” and a “risk” treatment in one study. Nevertheless, this has been done by Bonfim and Silva (2019) (On a sidenote: The reference list includes a wrong title in reference to Bonfim & Silva, 2019)). The way I see it, the main contribution of this paper really lies in the manipulation of punishments and therefore payoffs at the group versus the individual level. This contribution should be highlighted.  

4)        The effect of payoff differentials, as they result in the suggested individual punishment scenarios, on joint decision making has been previously investigated in the context of tax declarations. Relevant papers are Lohse and Simon (2021), as well as Matthaei and Kiesewetter (2022). These studies should be taken into account when assessing the contribution of the current study.

5)        Unfortunately, for me personally, the predictions are very hard to follow. I for one do not see how the expected payoff equals YL, or 0.5, in case of honest reporting according to Equation 1. Does Equation 1 truly represent the expected payoff for the individual or does Equation 1 represent the additional expected payoff in case of dishonest reporting? These two options are clearly not identical. I kindly ask the authors to clarify this. In addition, I suggest to state the expected payoff function in terms of true income, declared income, tax rate, audit and penalty rate as it is commonly done in tax compliance studies.

6)        Changing the wording and probability display from risk of punishment to probability of not-being punished decreases clarity and makes it harder to follow the explanations. I suggest to follow a common wording.

7)        Similarly, I do not understand how payoffs displayed in Table 1 are derived. From what I see in cell “Individual / 0% punishment risk”, Table 1 shows the expected additional payoff from being dishonest. Again, this is not the same as the overall expected payoff. Statements such as “In the individual payoff/no punishment treatment, the risk of being punished is zero and the gain from full dishonesty is £1.5, thereby resulting in an expected payoff of 1.5.” sadly do not help the confusion between “gain from dishonesty” and total “expected payoff”.

8)        As far as I understand the experimental setup the audit is always applied at the individual level. From the description of the experiment I see two different audit and punishment scenarios:

a.        Individual punishment: Each individual declaration is subject to an audit risk of 30% and only the individual that reported too little income is punished.

b.       Group punishment: Each individual declaration is subject to an audit risk of 30% but every member of the group is punished for every individual that reported too little income.

There is no treatment where only the group decision, i.e., the median of the individual decisions, is subject to an audit risk and penalties are based on the comparison of the group declaration to the true income. My understanding matches the statement on p. 17 that “Audits are individually applied across all treatments for each round.” However, when reading the experimental instructions (Supplementary Note 2 p. 8)  I see instructions which describe that only the group income will be compared to the true income in case of an audit. The only way I can see these instructions matching the description in the main text is if each individual declaration is replaced with the group declaration, i.e., the median individual declaration and then an individual audit is applied, in the group payoff/ group punishment treatment. If this is indeed the correct interpretation this has to be clarified in the main text but especially also in the experimental instructions. Otherwise, participants will not understand what will happen in the experiment.

9)        Related to my previous comment, participants cannot understand that there is an increased audit risk in case of group punishment given the current experimental instructions. This has to be made very explicit.

10)   Not including a treatment in which only the group decision is subject to an audit is a choice made by the experimenter. Nevertheless, this choice may be problematic for several reasons:

1)        It is not fully clear which treatment serves as a baseline comparison for the group payoff/ group punishment treatment. This treatment seems to differ in more than one aspect from the other treatments as

a.        The individual’s declaration has externalities as the median choice determines the individual’s payoffs.

b.       The individual being audited has externalities as all group members will have to pay a fine.

c.        The audit risk increases for the individual.

In my understanding, it is not possible to differentiate which of the three changes, A,  B, or C, causes a change in the individual reporting decision. Nevertheless, this may be related to my trouble to fully understand the experimental setup.

2)        The design is odd given that in reality the group decision is what would be reported to financial authorities and subject to an audit risk not the underlying individual decisions. Especially, the scenario of group payoff but individual punishment feels odd given this background. As far as I understand the experimental setup in this treatment, the individually reported income will be replaced by the median of the individual declarations and then the individual faces an audit risk for this declaration. This refers to a situation were someone else makes a decision but the subject faces the risk of this decision alone, e.g., in case of joint tax reporting partner A fills in the tax return but partner B has to pay the fine in case of an audit. This is possible for sure but seems rather constructed.

3)        The design choice further limits comparability to previous studies. This choice should be motivated and explained in more detail. The explanation currently included on page  17-18 is not convincing given the tax framing of the study.  

11)   The experimental setup is complicated and participants are required to do quite a bit of math to understand the outcome of their decision. Therefore, I suggest to include additional examples in the instructions. Could it be easier for the participants to abandon the 20% share of group income and to simply split income by 3? I was wondering if the 20%-share may be confused by participants with the 1 out 5 decisions that is chosen for payoff in the end.

12)   The demographics are not mentioned in the analysis plan. Are these going to be added as control variables in the regressions? I specifically wonder about asking for nationality – do you plan to investigate differences regarding heterogeneity between groups? I do not see a control for individual risk preferences. I suggest to add this.

References

Alm, J., & Malézieux, A. (2021). 40 years of tax evasion games: A meta-analysis. Experimental Economics, 24(3), 699–750.

Bonfim, M. P., & Silva, C. A. T. (2019). Inhibitory Mechanisms on Dishonesty of Groups and Individuals. Journal of Accounting, Management and Governance, 22(2), 205-226.

Fochmann, M., Fochmann, N., Kocher, M. G., & Müller, N. (2021). Dishonesty and risk-taking: Compliance decisions of individuals and groups. Journal of Economic Behavior & Organization, 185, 250-286.

Lohse, T., & Simon, S. A. (2021). Compliance in teams – Implications of joint decisions and shared consequences. Journal of Behavioral and Experimental Economics, 94, 101745.

Matthaei, E., & Kiesewetter, D. (2022). Group Decisions and Asymmetric Payoffs: Risky Tax Avoidance in the Laboratory. Available at SSRN: https://ssrn.com/abstract=3626982

Dear Dr. Rima-Maria Rahal, 

Thanks for giving me the opportunity to review this manuscript. I enjoyed reading the stage 1 report as well as other supplementary materials provided by the authors. Overall, my opinion is that the manuscript is well written and the experimental design clearly described. I believe that this work has the potential of being recommended by PCI RR after addressing a few remaining issues. Therefore, I suggest a revision for the current submission.

I am not as an expert on coding as the authors so in my review I have focused on the manuscript. I therefore leave code checking to other reviewers. With my limited training and exposure in behavioral economics, I did not find anything erroneous in the reasoning. But if you deem it necessary, finding an expert in this area would also be helpful to the authors. I asked colleagues with this background to co-review but unfortunately their schedule did not permit. 

Please see attached for my review report. 

Best,

Yikang