Insights / Research BriefMay 11, 2023

Toward an Understanding of Tax Amnesties: Theory and Evidence From a Natural Field Experiment

Patricia Gil, Justin E. Holz, John A. List, Andrew Simon, Alejandro Zentner
Receiving a message about a tax amnesty program (especially deterrence messages) increases the likelihood of joining the amnesty and repaying some debt; importantly, such amnesties do not cause a reduction in tax payments over the following two years, contrary to existing theoretical predictions.

As federal, state, and local budgets continue to tighten, governments have turned to an age-old approach to increase tax revenues: tax amnesties, or programs allowing taxpayers to voluntarily disclose unpaid taxes and to pay a defined amount, thus avoiding penalties. Since 2000, 84 countries have offered 184 amnesties, with 45 of those since 2012, including eight in the past two years. Amnesties are also popular among US state and local governments for not only taxes, but also issues ranging from human rights abuses and draft evasion to parking tickets and overdue library books.

While clearly popular, are amnesties effective? The theoretical literature suggests ambiguous effects in the short- and long-run, and many empirical questions remain unanswered. Likewise, amnesties have many detractors, especially surrounding taxes. Few tax debtors choose to participate in tax amnesties, detractors argue, likely because the costs of holding onto known and hidden debt is low; that is, governments often signal low enforcement capacity through amnesties, so why bother participating? Also, amnesties could worsen taxpayers’ perceptions of state capacity, further decreasing the incentive to join. With low take-up, amnesty opponents argue that there are minimal short-run benefits, and that offering evaders amnesty discourages future compliance, reducing long-run revenue. 

However, are these theoretical assumptions correct? Can a better empirical understanding reshape our theories? This authors address these and related questions by partnering with the Dominican Republic Tax Authorities (IRSDR) to implement a field experimental design via randomizing correspondences to taxpayers. In total, the IRSDR delivered messages to 125,452 debtors, who collectively owed $5.2 billion in known tax debt. The authors randomly allocated debtors to various treatments within four mediation channels (information, tax morale, expectations, and deterrence) to explore the drivers of amnesty take-up. Please see the working paper for more details on methodology, but this exercise reveals the following empirical insights:

  • Status quo take-up of the amnesty is relatively low. Only 18% of the eligible debtors join the tax amnesty in the control group. Overall, the government raises $263 million from the amnesty, just 5% of the total known debts owed. 
  • Each of the four messages increases the likelihood of paying both known and hidden debts relative to the control group. However, only the deterrence message highlighting the potential prison sentence increases the average debt paid. 
  • Firms respond more strongly to treatments than individuals, and larger taxpayers, based on the number of employees, respond more strongly than smaller ones.
  • Finally, and importantly, the authors find no evidence that subjects change their tax compliance in the subsequent year. In total, the messages induced treated debtors to collectively pay back $22 million more debt than control debtors, with no effect on subsequent income taxes paid.

Bottom line: Tax amnesties can work. This research provides clues about how taxpayers respond to tax amnesties and how governments can design effective programs. The authors show that receiving a message increases the likelihood of joining the amnesty, regardless of whether it contained behaviorally motivated language. Notably, the deterrence message highlighting potential prison sentences substantially increased both known and hidden tax debt repayments.