Firms as Tax Collectors (with Darío Tortarolo), Journal of Public Economics, May 2024, 233: 105092.

[Paper] [Working Paper] [World Bank Development Impact Blog]

We show that delegating tax collection duties to large firms can bolster tax capacity in weak-enforcement settings. We exploit two reforms in Argentina that dramatically expanded and subsequently reduced turnover tax withholding by firms. Combining firm-to-firm data with regression discontinuity and difference-in-differences methods centered on revenue eligibility thresholds, we find that: (i) appointing large firms as collection agents (CAs) does not hinder their economic activity, (ii) it leads to a significant increase in self-reported sales and tax payments among CAs' business partners, (iii) these effects are primarily concentrated among downstream firms that lack a traceable paper trail, and (iv) reductions in withholding lead to a decrease in self-reported sales, albeit to a lesser extent. Tax-collecting firms can thus help boost tax compliance and revenue.

Working Papers

Corporate Responses to Size-Based Tax Rates in LithuaniaWorld Bank Policy Research working paper n. WPS 10500
(with Thiago Scot)

This paper studies how firms respond to differential, size-based tax rates using administrative tax data in Lithuania. Exploiting a notch in the tax schedule faced by corporations, it documents strong behavioral responses to tax incentives—revenue elasticity is estimated at 0.35 and cost elasticity at -1.3, implying a large total profit elasticity of 7.4. It then leverages the panel structure of the data to provide insights on the dynamic effects of these tax incentives. Firms located close to but below the notch report systematically lower revenue growth in the short term, but the effects dissipate over time.

The Impact of COVID-19 on Formal Firms: Lessons from Administrative Tax DataRevise & Resubmit, Journal of Development Economics (with Pierre Bachas, Anne Brockmeyer, Camille Semelet)

Most low-income countries lack high-frequency firm-level data to monitor the effect of economic shocks in real time. We examine whether administrative tax data can help fill this gap, in the context of the COVID-19 pandemic. In spring 2020, we used the full population of corporate tax returns for 2019 in six developing countries to predict the effect of COVID-induced shocks on formal firms' activity. Comparing the predictions to the realized 2020 data, we find that firms were more resilient than predicted: the share of unprofitable firms increased by only 7 percentage points, while aggregate profits and taxes paid remained stable. The simulations failed to anticipate that labor and capital inputs would flexibly adjust and that large firms would be very resilient. Complementing our simulations with higher-frequency VAT data would have markedly improved predictions.

Robots, Exports and Top Income Inequality: Evidence for the U.S.CEDLAS working paper n. 0307
(with Andrés César, Guillermo Falcone)

The last decades have witnessed a revolution in manufacturing production characterized by increasing technology adoption and a strong expansion of international trade. Simultaneously, the income distribution has exhibited both polarization and concentration among the richest. Combining datasets from the U.S. Census Bureau, the U.S. Internal Revenue Service, the International Federation of Robotics, EU KLEMS, and COMTRADE, we study the causal effect of industrial automation on income inequality in the U.S. during 2010–2015. We exploit spatial and time variations in exposure to robots arising from past differences in industry specialization across U.S. metropolitan areas and the evolution of robot adoption across industries. We document a robust positive impact of robotics on income for only the top 1 percent of taxpayers, which is largest for top income fractiles. Therefore, industrial automation fuels income inequality and, particularly, top income inequality. According to our estimates, one more robot per thousand workers results in relative increments of the total taxable income accruing to fractiles P99 to P99.9, P99.9 to P99.99 and P99.99 to P100, of 2.1 percent, 3.5 percent and 5.9 percent, respectively. We also find that robotization leads to increased exports to high-income and upper-middle-income economies, and that this is one of the key mechanisms behind the surge in top income inequality.

Tax Filings vs Tax Withholdings: Behavioral Responses from Firms and Tax Preparers
(with Darío Tortarolo)

We study the relationship between tax filings and tax withholdings for the Argentine Turnover Tax on businesses and we shed light on the role of tax preparers in that relationship. Firms file a tax return every month through a centralized online platform and can (i) claim a refund if the amount withheld in advance (third-party reported by suppliers and clients) exceeds the tax liability, or (ii) remit the difference if the tax liability is higher than withholdings. Using rich administrative data, we first document sharp bunching exactly at the point where tax liabilities (reported by the firm) equal tax withholdings (third-party reported) and argue that this pattern is due to strategic under-reporting of income so that the tax balance matches the withholding shown by the online platform. Second, by linking firms to tax preparers through shared contact information we study the role of accountants on this behavior. We analyze how correlated are the behaviors of firms who share the same accountant (network). We find a positive relationship between accountants and the bunching behavior observed on their clients. We also exploit information on past audits and show that, within a network, non-audited firms respond to their peer's audit, suggesting that tax preparers might act as diffusers of information across firms.

Work in Progress

Payments Under the Table in Latin America (with Javier Feinmann, Ana Paula Franco, Nathalie Gonzalez-Prieto, Roberto Hsu, Maximiliano Lauletta)

This paper investigates a neglected aspect of informality in Latin America—Payments Under the Table (PUT), where registered firms make off-the-books salary payments to registered employees. We conduct the first multi-country large-scale survey on this topic, covering Argentina, Brazil, Chile, Colombia, and Mexico. Out of the more than 20,000 formal workers surveyed, our results indicate that 17% of them receive some part of their compensation under the table. Among PUT receivers, on average, 24% of their labor earnings are paid off the books. We then provide insights into the mechanics and motivations behind PUT and analyze policy implications. By shedding light on this overlooked dimension, the study contributes to a more comprehensive understanding of informality's role in Latin American economic development.

Firm Dynamics Along the Development Path
(with Pierre Bachas and Anne Brockmeyer)

We study firm-growth dynamics in developing countries and its macroeconomic implications. This area of study has been hampered by data limitations since most countries have only occasional manufacturing surveys or censuses at best. Consequently, while the literature has documented the existence of market distortions leading to limited firm growth, this evidence has been restricted to a small set of countries that have available data. In turn, this has also limited the possibility of gaining insights at an aggregate level. In this project, we harmonize and combine administrative tax data for multiple countries. This allows us to trace out within-country trends as well as gaining knowledge of firm dynamics along different stages of development.

Assessing the role of tax preparers in tax compliance: direct and network effects
(with Darío Tortarolo)

  • Paused due to COVID-19

Two understudied areas in Public Finance are the role of tax professionals and the role of networks in tax administration and enforcement (Slemrod, 2018). In this project we seek to answer two questions: First, whether it is more cost-effective to communicate tax preparers or taxpayers to improve tax compliance; Second, whether there are spillover effects from targeted to non-targeted taxpayers that form part of the same network (those sharing the same accountant). To that end, we run a large-scale randomized communication experiment where we send deterrence emails to taxpayers and/or accountants (about 100,000 taxpayers in the treatment group and 900,000 in the control group).


“Exports of Argentina and Brazil under the Generalized System of Preferences” (with Facundo Albornoz and Irene Brambilla). Ensayos Económicos, 1(74), pp. 27-55 (2016).