Estimating Income in a Tax Compliance Game. A Bayesian Persuasion Approach

This paper studies the tax authority’s problem of how to estimate a tax payer’s income in a tax compliance game. The tax authority’s choice of how to estimate income is modelled using the Bayesian persuasion framework and assuming that income can be estimated arbitrarily precisely. I show that the tax authority can use income estimates as a commitment device: ex-post, the tax authority has an incentive to audit a tax payer if his income is estimated to be high. This allows the tax authority to increase tax compliance by strategically overestimating low income. If the probability with which low income is falsely estimated to be high is strictly positive, the tax authority audits a low income tax payer with a strictly positive probability. Anticipating to be audited with a sufficiently high probability, the tax payer prefers to report low and high income to avoid being audited and fined.