Discretion versus algorithms: bureaucrats, tax equity and acceptability

We study how replacing bureaucrats’ discretion with algorithmic assessment affects the accuracy, equity, and public acceptance of tax decisions. Using Senegal’s first digital property tax census in the capital, we experimentally vary reliance on bureaucrats versus algorithms to value properties, and compare results to benchmark values provided by real estate assessors. Algorithms significantly improve accuracy and both horizontal and vertical tax equity, with fully rule-based systems outperforming hybrid methods. Performance gaps reflect bureaucrats’ limited knowledge rather than collusion. Conditional on liability, taxpayers are indifferent to algorithmic decisions, making algorithms a promising tool to expand the tax base without political backlash.