The effect of corruption on tax evasion

Last registered on May 18, 2022

Pre-Trial

Trial Information

General Information

Title
The effect of corruption on tax evasion
RCT ID
AEARCTR-0009322
Initial registration date
May 17, 2022

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
May 18, 2022, 5:12 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information

Primary Investigator

Affiliation
University of Nottingham

Other Primary Investigator(s)

PI Affiliation
WZB Berlin Social Science Center

Additional Trial Information

Status
In development
Start date
2022-05-16
End date
2022-09-16
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Previous research on tax compliance has shown that individuals’ decision to pay taxes is shaped not only by intrinsic motivation but also by extrinsic factors like perceived fairness of the tax system, reliability of the political system, and, more specifically, the perceived level of corruption in the society. In this study, we examine whether an environment of widespread corruption makes taxpayers evade more taxes. We plan to implement treatment variations to disentangle whether the effect of corruption on tax evasion is explained by the damage to the public good that corruption generates or just by having a corrupt administration in place. We divide our study into two experiments. The first one is a Public Officials experiment used as input to create an objective prior about the level of corruption by collecting Public Officials' decisions of misappropriation. The second experiment is a tax evasion game where participants can lie about their income after learning the level of corruption in their group.
External Link(s)

Registration Citation

Citation
Parra, Daniel and Yuliet Verbel. 2022. "The effect of corruption on tax evasion." AEA RCT Registry. May 18. https://doi.org/10.1257/rct.9322
Sponsors & Partners

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details

Interventions

Intervention(s)
We study whether an environment with a high level of corruption influences tax evasion and whether the loss of resources from the public good can explain that effect if any.
Intervention Start Date
2022-05-16
Intervention End Date
2022-09-16

Primary Outcomes

Primary Outcomes (end points)
Evasion Rate = 1 - reported income/actual income
Evasion Size = actual income - reported income
Evasion dummy = 1 if Evasion Size>0

Note: in the case of Evasion Size < 0 we will consider this as overreporting and will exclude the corresponding observations before the data analysis.
Primary Outcomes (explanation)
Our analysis will be solely based on the decision of Tax Payers. Participants under the role of Tax Payers earn an income and have to report it for tax payment purposes. We compare their actual income to their income report. We understand tax evasion as underreporting income for tax payment purposes.

Secondary Outcomes

Secondary Outcomes (end points)
Appropriateness rate of different income report decisions for tax purposes.
Secondary Outcomes (explanation)
We plan to conduct a norm elicitation questionnaire (Krupka and Weber, 2013) asking for a rating of the appropriateness of different income report decisions for tax purposes. We will use this information to check whether participants' income decisions are related to how appropriate they consider tax evasion in each treatment and to understand the general perception of how appropriate tax evasion is among the population we are eliciting decisions from.

Experimental Design

Experimental Design
We study whether corruption influences tax evasion and whether the loss of resources in the public good can explain that effect if any. For this purpose, we plan to implement a tax evasion game with groups of four participants: three taking the role of taxpayers and one taking the role of official. All participants earn money in an initial stage where they perform an encoding task for 5 minutes. The taxpayers can lie about their earned income, and the official can take money from the collected taxes. In order to identify the causal effect of corruption on tax evasion, taxpayers will be informed about the level of corruption in their group before they make the reporting decision. The money collected for taxes is augmented by an efficiency factor and is transferred to the organization Keep Britain Tidy which is a real public good for UK residents. If an official decides to take money from the collected taxes, that generates a loss of resources in the public good.

In the case of the officials, after they earn their income, they decide whether they want to take a fixed amount of £0.63 from the taxes collected. We gather the taking decisions of the officials and create two groups, one with low corruption (fewer officials in the group that decided to take) and one with high corruption (majority of officials in the group that decided to take). In the case of the taxpayers, after they earn their income, they are informed about the type of group where the official they are matched with comes from (low/high corruption), and then they are asked to report their income for taxation. The income reported is taxed at a rate of 35%.

We plan to implement treatment variations to disentangle whether the effect of corruption on tax evasion, if any, is explained by the loss of resources in the public good that corruption generates or just by having a corrupt administration in place. For this, we will implement two more treatments (low/high lottery) where the officials instead of actively deciding whether to take the £0.63, a lottery will randomly determine whether they win £0.63 or not. In this sense, the same loss of resources occurs but the cause differs in comparison to the treatments where they actively decided whether to take the money.

Final payments are as follows: officials’ payoff is their earned income plus £0.63 if they decide to take it or if the lottery assigns it to them, otherwise, the payoff is only their earned income. And taxpayers’ payoff is their earned income minus 35% of their reported income.

The experiment will be neutrally framed, which implies that participants’ roles will be labeled as Participant A for officials and Participant B for taxpayers.

From hereafter, the names of the treatments where there are high/low levels of corruption given the decision to take of the officials will be Choice-High and Choice-Low. The names of the treatments where a lottery determines if an official takes or not will be Random-High and Random-Low.

Hypotheses
H1: The evasion rate will be higher in Choice-High than in Choice-Low.
H2: The evasion rate will be higher in Choice-High than in Random-High.
H3: The evasion rate will be higher in Random-High than in Random-Low.
H4: The evasion rate will be equal in Choice-Low than in Random-Low.

Secondary Hypothesis on appropriateness rates
H1: Individuals will consider it more socially appropriate to evade taxes in Choice-High than in Choice-Low.
H2: Individuals will consider it more socially appropriate to evade taxes in Choice-High than in Random-High.
H3: Individuals will consider it more socially appropriate to evade taxes in Random-High than in Random-Low.
H4: Individuals will consider it equally socially appropriate to evade taxes in Choice-Low than in Random-Low.

Note: In version 1 of this pre-registration we do not know the exact distribution of the Public officials for High/Low treatments. The distribution will depend on the decisions of Public officials in the Choice treatment.
Experimental Design Details
Not available
Randomization Method
We will randomly assign treatments ex-ante at the individual level.
Randomization Unit
Individual-level.
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
200 groups of four participants. Each group with one Public Official and three Tax Payers.
Sample size: planned number of observations
800 participants in total from which we obtain 600 independent observations of income reporting.
Sample size (or number of clusters) by treatment arms
50 Public Officials + 150 Tax Payers in Choice-High
50 Public Officials + 150 Tax Payers in Choice-Low
50 Public Officials + 150 Tax Payers in Random-High
50 Public Officials + 150 Tax Payers in Random-Low
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
We calculated the power using Monte Carlo simulations. We used a minimum detectable effect size of 0.1pp on the Evasion Rate from people detected underreporting their income for tax payment purposes. The power reached with the sample size of 150 observations per treatment is higher than 0.8 when simulating 1000 t-tests.
IRB

Institutional Review Boards (IRBs)

IRB Name
WZB Research Ethics Review
IRB Approval Date
2021-09-10
IRB Approval Number
2021/3/125