We experimentally test the effect of aggressive audit strategies on tax compliance. Taxpayers first go through a phase of audits managed by a human tax agent who is requested to follow a rule imposed by a fair random device. However, the tax agent can freely decide to break the rule and over-inspect. Afterward, taxpayers are exposed to a genuinely random audit process governed by an algorithm, which makes compliance a strategically dominated option. We find that taxpayers are generally over-inspected by the human tax agents and react to this with nearly full compliance. Our main result is that these high levels of compliance also persist when controls are implemented by the algorithm. This suggests that tax authorities can use aggressive audit strategies to raise and sustain tax compliance.

我们通过实验测试了积极的审计策略对税收合规性的影响。纳税人首先要经过由人力税务代理人管理的审计阶段，该审计员应遵循公平随机设备施加的规则。但是，税务代理可以自由决定违反规则和进行过度检查。之后，纳税人将面临由算法控制的真正随机审计流程，这使合规性成为策略性主导的选择。我们发现，纳税人通常会受到人力税代理人的过度检查，并对此做出几乎完全合规的反应。我们的主要结果是，当算法实现控制时，这些高水平的合规性也将持续存在。这表明税务机关可以使用积极的审计策略来提高和维持税收合规性。