Abstract:
Many countries grant exemption from legal prosecution under certain conditions, allowing for voluntary disclosures regarding tax evasion. Empirically, tax amnesties appear most successful when accompanied by an increase in enforcement efforts because they help evaders to adjust to the new circumstances. Time-limited amnesties are often repeated or in some countries even permanent. Anticipated tax amnesties can, however, serve as an insurance against a rise in the detection probability, potentially leading to less and not more tax compliance. We test the relevance of this insurance effect in an experimental tax game and disentangle it from an effect on tax morale, i.e., non-pecuniary motives to comply. We ftnd that the former effect indeed decreases overall tax compliance by about 7-10 percent, showing its relevance for the design of tax policies. A high predictability of amnesties may work to their detriment.