Bridging tax compliance gaps

A two-day workshop by ATI and UNU-WIDER on practical applications of the Tax Gap Estimation Toolkit

The Tax Gap Estimation Toolkit, developed by the United Nations University World Institute for Development Economics Research team with support from the Addis Tax Initiative (ATI), is an innovative tool designed to help countries estimate compliance gaps and integrate these estimates into their compliance risk management strategies. By doing so, tax authorities can better allocate resources and improve audit selection processes. 

On 7 and 8 October 2024, over 90 tax experts and practitioners participated in a two-day training session on the toolkit. Delivered by UNU-WIDER researchers, the training provided participants with practical skills to estimate tax compliance gaps using a bottom-up approach that incorporates machine learning, aiming to enhance tax revenue mobilisation efforts. 

Screenshot taken during the presentation of the two-day training program
Screenshot taken during the presentation of the two-day training programme.

"The training is designed to support policymakers and tax authorities in identifying and quantifying gaps in tax revenue collection. The toolkit is intended to benefit colleagues who are directly involved with estimation and analytical tasks in revenue authorities”.

Day 1: Bottom-up approach and data preparation

Dr. Amina Ebrahim, Research Fellow at UNU-WIDER, kicked off Day 1 with a detailed breakdown of the bottom-up approach that underpins the Tax Gap Estimation Toolkit. She noted how this method leverages microdata from tax audits, making it adaptable to various country contexts, particularly those using risk-based or operational audits.

In her presentation, Dr. Ebrahim also revisited the key discussions from the ATI Tax Gap workshop held in Dar es Salaam in March 2024, explaining how the toolkit builds on methodologies examined during that event. The workshop, attended by experts from the World Bank, IMF, and tax authorities from all member countries, laid the groundwork for the development of this toolkit. It went beyond simply emphasising the need for accurate tax gap estimations, offering practical methodologies taught by the IMF, UNU-WIDER, and the World Bank. Participants engaged in hands-on sessions, taking steps towards closing compliance gaps and improving tax revenue mobilisation by applying these methods.

Pictures taken during the ATI Tax Gap Workshop in Tanzania, March 2024
Pictures taken during the ATI Tax Gap Workshop in Tanzania, March 2024.

 

Dr. Ebrahim stressed the toolkit's relevance to ATI Commitment 1, which is outlined in the ATI Declaration 2025, focusing on enhancing the effectiveness of revenue administrations.

"The development of this toolkit is motivated by its potential to fulfil ATI Commitment 1, as it strengthens risk management frameworks and improves non-compliance strategies".

Slide on ATI commitment 1
Slide on ATI Commitment 1. 

 

Data preparation and practical applications

Following Dr. Ebrahim’s comprehensive introduction, Dr. Maria Jouste, Research Associate at UNU-WIDER, led the session on data preparation. She outlined the importance of cleaning and managing tax administration data to ensure accurate estimates. Dr. Jouste emphasised that the effectiveness of any tax gap analysis relies heavily on the quality and thorough preparation of the data used.
Participants were then divided into breakout rooms for practical sessions where they could apply the concepts discussed, work with real data sets, and put the toolkit into practice.

Dr. Maria Jouste, Research Associate at UNU-WIDER leading a session on data preparation and practical applications
Dr. Maria Jouste, Research Associate at UNU-WIDER leading a session on data preparation and practical applications.

 

Dr. Ebrahim highlighted the interactive nature of the training:

"These practical sessions are critical because they allow participants to test the toolkit themselves, ask questions, and explore how it can be tailored to their country’s data systems".

Meaningful engagement

A question was raised during the Q&A session regarding the use of machine learning models in comparison to traditional methods. Dr. Ebrahim provided an insightful response, explaining that while machine learning isn’t necessarily superior to all models, it performed better in the cases tested, particularly in predicting tax outcomes compared to audit results.

Q&A session on the use of machine learning models
Q&A session on the use of machine learning models.

 

During the same session, a concern was raised about the quality of tax administration data and how this might affect the application of the bottom-up approach. The discussion highlighted challenges faced by tax administrations with data quality and the limitations of audit data, especially for those with limited resources to conduct comprehensive audits. Dr. Ebrahim responded by offering practical advice on how to work with imperfect data, emphasising the importance of focusing on usable data while recognising its limitations.

Mr. Alex Oguso from the Kenya Revenue Authority (KRA)
Mr. Alex Oguso from the Kenya Revenue Authority (KRA).

 

Dr. Maria Jouste followed up, sharing insights from her experience in Uganda:

"In Uganda, we’ve explored using assessment data instead of audit data, which has proven useful for estimating tax gaps when audit data is incomplete. We’re currently testing this and will share the results soon."

Day 2: Estimation theory and practical application

Day 2 opened with Dr. Kwabena Adu-Ababio, Research Associate at UNU-WIDER, who delved deeper into estimation theory. His presentation focused on using machine learning to estimate tax gaps, particularly for non-audited firms. This method helps to predict tax outcomes for firms that haven’t been audited, thereby providing a fuller picture of the compliance gap.

"We aim to use machine learning to predict potential revenue for non-audited firms based on the patterns identified in audited data. This helps us better estimate the compliance gap, which is our main focus."

Dr. Kwabena Adu-Ababio

The second day also featured practical breakout sessions where participants applied estimation techniques to real data sets. This hands-on approach allowed them to gain deeper insights into how the toolkit can be effectively used.

Dr. Kwabena Adu-Ababio, Research Associate at UNU-WIDER, presenting the process of calculating and plotting the tax gap during the ATI Tax Gap Estimation Toolkit training
Dr. Kwabena Adu-Ababio, Research Associate at UNU-WIDER, presenting the process of calculating and plotting the tax gap during the Tax Gap Estimation Toolkit training.

Concluding thoughts and future applications

As the training concluded, Dr. Amina Ebrahim offered a recap of the two-day session, reiterating the importance of applying the toolkit to enhance tax compliance strategies:

"The toolkit is designed to support countries in estimating compliance gaps and integrating these estimates into their compliance risk management strategies. This can help authorities prioritise resources and improve audit selection processes".

The Tax Gap Estimation Toolkit is set to officially launch at the end of 2024, with feedback and insights from the training being incorporated into the final version. Additional practical sessions will also be held until the end of this year to support participants in applying the toolkit effectively.

Final Recap and Application of Tax Gap Estimation Results
Final recap and application of Tax Gap estimation results.

The training underscored the value of collaborative learning and highlighted how this toolkit, once launched, will provide a valuable resource for tax authorities worldwide as they work to close compliance gaps and improve tax revenue mobilisation.

Group Photo: ATI Tax Gap Estimation Toolkit Training
Group photo: Tax Gap Estimation Toolkit training.