Publications

Peer-reviewed publications, listed in reverse chronological order.

Where Do My Tax Dollars Go? Tax Morale Effects of Perceived Government Spending

with Brad Nathan, Ricardo Perez-Truglia, and Alejandro ZentnerAmerican Economic Journal: Applied Economics, 2025, 17(4): 223–259

Expand

This paper asks whether people’s willingness to pay taxes depends on what they think their taxes are used for. Using a real-world experiment, we show that correcting taxpayers’ mistaken beliefs about government spending changes their behavior, for example, whether they appeal a property tax assessment. The results suggest that people are more willing to pay taxes when they learn that those taxes fund public services that are more valuable to them.

Figure showing the estimated effect of information about the school-tax share on the probability of filing a property-tax appeal, separately for households with and without children. The figure reports both unadjusted estimates and estimates controlling for baseline characteristics. The information treatment reduces appeal filing by about 4.8 percentage points among households with children and increases it by about 4.8 percentage points among households without children. The difference between the two groups is statistically significant, and the adjusted and unadjusted estimates are similar.
Average Treatment Effects of Feedback on Share of Property Taxes that Go to Public Schools on Probability of Protesting Taxes.

Tax Audits as Scarecrows: Evidence from a Large-Scale Field Experiment

with Marcelo Bérgolo, Rodrigo Ceni, Guillermo Cruces, and Ricardo Perez-TrugliaAmerican Economic Journal: Economic Policy, 2023, 15(1): 110–153

Expand

This paper asks whether firms evade taxes by weighing the gains from evasion against the risk and cost of being caught. In collaboration with the Uruguayan tax authority, we conducted a real-world experiment involving 20,440 small- and medium-sized firms that together paid more than US$200 million in taxes each year. We find that providing firms with information about audits changed their tax compliance decisions, but not in the way predicted by the standard model of tax evasion.

Figure showing the estimated effect of an audit-statistics message on tax compliance across deciles of the audit-probability signal included in the letter, ranging from 2 to 25 percent. The estimated effects are similar across the full range of signals, with no evidence that higher stated audit probabilities produce larger compliance responses. A dashed linear fit is nearly flat, indicating no statistically significant relationship between the signal value and the treatment effect.
Treatment effect of the probability of being audited on VAT payments.

Dissecting Inequality-Averse Preferences

with Marcelo Bérgolo, Gabriel Burdín, Santiago Burone, Mauricio De Rosa, and Martín LeitesJournal of Economic Behavior & Organization, 2022, 200: 782–802

Expand

This paper asks what shapes individuals’ inequality aversion. We surveyed more than 1,800 students in Uruguay and asked them to choose between societies with different levels of average income and inequality. We also varied the source of inequality, the income position participants were asked to consider, and opportunities for social mobility. Most were willing to sacrifice some income for greater equality, but less so when inequality resulted from effort rather than luck. Highlighting opportunities for social mobility had different effects depending on the income position considered: it reduced concern about inequality when mobility would move them up, but increased it when mobility would move them down.

Figure showing estimated treatment effects on inequality aversion by the imagined grandchild’s position in the income distribution: minimum, mean, or maximum. Dots indicate point estimates and bars show 95 percent confidence intervals. The effects of the effort and luck messages are similar across all three positions. The mobility treatment has a negative effect when the grandchild is at the bottom, no detectable effect at the mean, and a positive effect when the grandchild is at the top.
Treatment Effects on Inequality Aversion by Position in the Income Distribution and for Different Treatment Arms.

Digging into the Channels of Bunching: Evidence from the Uruguayan Income Tax

with Marcelo Bérgolo, Gabriel Burdín, Mauricio De Rosa, and Martín LeitesThe Economic Journal, 2021, 131(639): 2726–2762

Expand

This paper asks how individuals respond to personal income taxes in Uruguay. Using detailed tax records, we examine taxpayers close to the point where the tax rate increases. We find that the overall response is modest and does not come from earning less. Instead, some taxpayers claim more deductions and, in some cases, underreport income. The findings suggest that taxpayers’ responses depend strongly on the opportunities the tax system gives them to reduce their taxable income. In Uruguay, where most workers are wage earners and deductions are limited, these opportunities, and therefore the overall response, remain relatively small.

Bunching graph for taxpayers who claim itemised deductions. The distribution shows a pronounced hump in taxable income around the kink point, with substantially more taxpayers than predicted by the counterfactual distribution. The estimated excess mass is 75 percent, corresponding to an elasticity of taxable income of 0.145.
Taxable labour Income Bunching: Pure Wage Earners who Itemize Deductions.

Misperceptions about Tax Audits

with Marcelo Bérgolo, Rodrigo Ceni, Guillermo Cruces, and Ricardo Perez-TrugliaAEA Papers and Proceedings, 2018, 108: 83–87

Expand

This paper asks what firms know about tax audits. Using a survey of Uruguayan firms and tax records, we find that firms greatly overestimate how likely they are to be audited, although their beliefs about penalties are much closer to reality. These mistakes are common even among accountants and experienced firms, and seem to be shaped mainly by whether a firm was audited recently.

Histogram showing firms’ perceived probability of being audited. Most respondents report probabilities above the actual audit rate of 7.98 percent, marked by a vertical line. The average perceived audit probability is 39.5 percent, indicating substantial overestimation of audit risk.
Probability of being audited: True vs Perceived.