Essays on Social Assistance and Tax Administration in Selected Developing Economies

In this doctoral dissertation, there is a collection of an introductory chapter and three independent studies on the fields of development and public economics. I discuss issues relating to fiscal policy and how households and firms are affected depending on the policy at play in some emerging economies. For households, I consider the extent to which (non)existing safeguarding mechanisms contribute to their welfare. For firms, I consider specific tax administrative interventions that affect behavior and domestic revenue mobilization. Using rich household and administrative data, I study how households and firms respond to shocks as well as tax reforms and audits respectively.

The first essay uses mainly microsimulation approaches to study the policy elements that can be implemented to limit the impact of a systemic shock. The choice of microsimulation stems from the ability of such models to conduct micro-level analysis that singles out the exact role of tax-benefit policies in specific environments where they operate. Although such techniques have been used in a few studies evaluating shock impacts in developed countries, examining several scholarly works reveals that similar analysis still needs to be improved for transitioning economies. The study takes as given the redistributive preferences of policy and tries to ascertain how the poor are or will be protected when there is a marginal (5%) income or demand shock based on the magnitude and composition of automatic stabilization coefficients for Ghana, South Africa, and Ecuador. Moreover, this exercise also serves as a stress test of the existing benefit system. I measure a household’s exposure to risk when there is a crisis with or without government intervention. The study also simulates ways to study how discretionary action has served as welfare contingency to cushion against random shocks statically. The microsimulation approach permits the investigation into the morning-after effects of different types of shocks on household disposable income, holding everything else constant to single out the exact role that automatic stabilizers play.

The second essay, co-authored with Aliisa Koivisto and Andreya Kumwenda, studies the impact of an administrative tax reform by the Zambian tax authority who, due to informality and poor bookkeeping, changes the reporting mechanism of the Value Added Tax (VAT). The reform is implemented when the Zambian Revenue Authority appoints relatively compliant and public firms as Withholding Agents who collect VAT when purchasing inputs from suppliers. The latter may include firms who were a priori non-compliant or initially compliant but vanished at the end of the transaction chain, exacerbating avoidance or evasion tendencies. The so-called Withholding Value Added Tax (WVAT) reform in practice reduces the lapses which are likely to occur predominantly in developing countries where informal suppliers are usually not found in tax registers. Using a difference-in-differences approach, the study estimates the impact of the WVAT reform implemented to improve revenue generation from suppliers in Zambia on value-added, sales, purchases, and output VAT. After the reform, there are significant positive impacts on those indicators that the withholding agent reports and remits to the tax authority. The study finds a coarse estimated increase in VAT revenue indicating improved compliance as remitting liability changes.

The third essay, co-authored with Aliisa Koivisto and Evaristo Mwale, studies domestic revenue loss through deviations in the potential and actual taxes raised by the Zambian Revenue Authority (ZRA) from firm reported VAT and CIT payments. The bottom-up approach is used to gather estimates as opposed to the top-down approach due to data constraints. The approach permits the use of tax assessments after audits which we classify as evasion rates and the reported returns of the compliant taxpayer population. Before the proposed approach is implemented, there is the need to estimate tax assessments for unaudited firms as the ZRA, like other revenue authorities, does not audit all firms but selects a sub-sample of the taxpaying firms for scrutiny assessments. The study, therefore, tries to predict evasion rates for the unaudited cohort of firms using a Regression and a Machine Learning approach based on the characteristics of audited firms prompting selection into audits. The regression output is based on a standard panel model and estimates total tax gaps at 55%. In contrast, the Machine Learning output is based on an Artificial Neural Network algorithm and assesses total tax gaps at 47%. For deviations in tax type, the study observes that corporate taxes mainly drive the differences. In critical sectors of the economy, the extractives sector records the highest value of CIT gaps compared to VAT gaps.

Contact Kwabena Adu-Ababio

Email: kwabena.adu-ababio@helsinki.fi
LinkedIn: https://www.linkedin.com/in/kwabena-adu-ababio/