Tax Evasion in Africa and Latin America : The Role of Distortionary Infrastructures and Policies

This paper examines the impact of the quality of the business environment as well as the monitoring capacity of the tax agency on firms' tax evasion and production decisions. First, the paper uses firm-level data for 30 African and Latin Ameri...

Full description

Bibliographic Details
Main Authors: Kouame, Wilfried A., Goyette, Jonathan
Format: Working Paper
Language:English
Published: World Bank, Washington, DC 2018
Subjects:
Online Access:http://documents.worldbank.org/curated/en/937371531333408137/Tax-Evasion-in-Africa-and-Latin-America-the-role-of-distortionary-infrastructures-and-policies
http://hdl.handle.net/10986/29995
Description
Summary:This paper examines the impact of the quality of the business environment as well as the monitoring capacity of the tax agency on firms' tax evasion and production decisions. First, the paper uses firm-level data for 30 African and Latin American countries to show that tax evasion and distortions stemming from the business environment are positively and significantly correlated, while sales not reported for tax purposes and institutional quality are negatively and significantly correlated. Second, the paper develops a general equilibrium model where heterogeneous firms make tax evasion decisions based on their assessment of the quality of their business environment as well as the monitoring capacity of the tax agency. The model simulations for each country in the African and Latin American sample show that the model can explain 35 percent of the variation in tax evasion and more than 49 percent of the dispersion in output per worker across the sample countries. Finally, a series of counterfactual experiments shows that, at the current level of deterrence, governments could decrease sales not reported for tax purposes by 21 percent, by reducing distortions stemming from the business environment by half. The paper presents empirical supporting evidence consistent with testable predictions of the model.