Experimental Design
Through the Common Reporting Standard (CRS) and the Foreign Account Tax Compliance Act (FATCA), the Danish tax authorities receives information from foreign banks about the financial assets and income of Danish taxpayers in foreign banks.
We measure, for each taxpayer, the potential underreporting of foreign dividends and interest income as the discrepancy between foreign income reported by banks under CRS and FATCA and the foreign income reported by the taxpayers themselves.
We stratify the sample of taxpayers into three groups by the size of their discrepancies:
• “Top group”: the 100 tax payers with the largest discrepancies
• “Middle group”: the 1000 tax payers outside the top group with the largest discrepancies
• “Bottom group”: All other tax payers with discrepancies larger than DKK 5,000
We randomly select 264 taxpayers from the “Middle group” and 136 taxpayers from the “Bottom group” for the audit intervention (Treatment group of 400 taxpayers). We randomly select the same number of taxpayers from each group for a control group, which is not audited (Control group of 400 taxpayers). In addition, we randomly select a buffer group of 66 taxpayers from the Middle group and 34 from the Bottom group (total of 100 taxpayers).
The tax authority should include taxpayers from this buffer group in the treatment group if the audits are less costly than expected. All 100 taxpayers in the “Top group” are selected for audit; however, as there is no comparable control group for these individuals, we will not use these observations when we estimate the effect of the intervention. Taxpayers in the middle and bottom group who are not selected for either the treatment group (including the buffer) or the control group may or may not be audited depending on the normal procedures of the tax authorities and we will not use these observations when we estimate the effect of the intervention. The number of taxpayers selected from each group is chosen to minimize the variance on the estimate of aggregate non-compliant income reported by foreign banks under CRS and FATCA (see below).
The Danish tax authorities conduct targeted audits focused on foreign interest and dividend income on the taxpayers in the Treatment group. The audits started on 4 October 2021. We expect them all to be initiated by 31 December 2021 and to be complete by 31 August 2022 although some uncertainty remains about these dates.
The goal of the experiment is to estimate the effect of the targeted audit on subsequent taxpayer behavior as observed on self-reported income and reports from foreign banks under CRS and FATCA in 2022. The main outcome reflects potentially non-compliant CRS/FATCA income:
• Discrepancy between foreign interest and interest income as reported by foreign banks under CRS/FATCA and self-reported on the tax return
We also investigate the effect on the following outcomes, which allows for a decomposition of the main outcome by income type and bank-reports vs self-reports:
• Discrepancy foreign interest income (CRS+FATCA reports and tax return)
• Self-reported foreign interest income (tax return)
• Bank-reported foreign interest income (CRS+FATCA reports)
• Discrepancy foreign dividend income (CRS+FATCA reports and tax return)
• Self-reported foreign dividend income (tax return)
• Bank -reported foreign dividend income (CRS+FATCA reports)
• Self-reported foreign dividend and interest income (tax return)
• Bank -reported foreign dividend and interest income (CRS+FATCA reports)
Finally, we will also consider the following secondary outcomes, which capture potential compliance spill-overs to other foreign income classes (e.g. capital gains) as well as the total tax consequences of the intervention:
• Self-reported foreign capital gains / losses (tax return)
• Self-reported total foreign capital income (tax return)
• Taxes paid (tax return)
As the sample is relatively small and the distribution of raw amounts is likely to be heavily skewed, we will consider outcomes expressed as raw amounts as well as alternative transformations:
• Logarithm / Inverse hyperbolic sine
• Indicators capturing the extensive margin (e.g. any foreign interest income)
We will also use the data collected in the context of the audits for other purposes. First, we will use it to estimate how much of the income reported by foreign banks under CRS and FATCA is non-compliant. We will do this by scaling up the non-compliant income found in the audits to population level by applying sampling weights. Second, for the cases where the audits find that the discrepancy between foreign income reported by banks under CRS and FATCA and the foreign income reported by the taxpayers do not reflect non-compliance, we will investigate the reasons for the discrepancy by conducting a survey for the auditors.