Tax revenue forecasting is important for government budgeting and fiscal planning. In Zimbabwe, forecasting tax revenue is difficult because the economy is unstable. The country experiences frequent inflation, currency changes, and policy reforms. These conditions make traditional forecasting models less reliable because they assume stable economic relationships.
This study develops an integrated econometric framework for forecasting tax revenue collected by the Zimbabwe Revenue Authority (ZIMRA). The research focuses on four major tax categories: Value Added Tax (VAT), Pay As You Earn (PAYE), Corporate Income Tax (CIT), and Customs Duties. Monthly data from January 2018 to December 2025 was used. The study combines three quantitative methods: Markov-switching models to detect different economic regimes, vine copula models to analyse relationships between tax revenues, and empirical Bayesian credibility methods to improve parameter estimation under uncertainty.
The results show that Zimbabwe’s tax revenues move between two economic regimes: stable and turbulent periods. Corporate Income Tax shows the highest volatility, while PAYE is more stable because it is based on wages. The dependence analysis indicates that most tax revenue streams are weakly related after accounting for macroeconomic conditions, although a negative relationship exists between corporate tax and customs duties.
Out-of-sample forecast evaluation shows that the baseline Markov-switching model produces slightly better point forecasts. However, the integrated framework generates more realistic forecast intervals and better captures revenue shocks.
At a glance
- Empirical analysis of monthly ZIMRA tax revenues from January 2018 to December 2025
- Evaluates four main revenue streams: VAT, PAYE, corporate income tax, and customs duties
- Combines Markov-switching regression with vine copulas and empirical Bayesian credibility adjustments
- 30 sources in the bibliography
- 6 figures
Table of Contents
1 INTRODUCTION
1.1 Introduction
1.2 Background of the study
1.2.1 Relevance of Tax Revenue Forecasting in Fiscal Management
1.2.2 Tax Revenue Forecasting Challenges in Developing Economies
1.2.3 Zimbabwe’s Macroeconomic Environment and Fiscal History
1.2.4 Conventional tax revenue forecasting approaches
1.2.5 Advanced tax revenue forecasting under the Integrated Modelling Framework for ZIMRA
1.3 Statement of the problem
1.4 Research Aim
1.5 Research objectives
1.6 Research questions
1.7 Research hypothesis
1.8 Significance of the study
1.9 Scope of the Study
1.10 Assumptions of the Study
1.11 Limitations of the Study
1.12 Organization of the Study
1.13 Chapter Summary
2 LITERATURE REVIEW
2.1 Introduction
2.2 Conceptual framework
2.3 Definition of key terms
2.3.1 Tax Revenue Forecasting
2.3.2 Regime-Switching Models
2.3.3 Markov-Switching Models
2.3.4 Vine Copulas
2.3.5 Empirical Bayesian Credibility
2.3.6 Macroeconomic Indicators
2.3.7 Tax Revenue Streams
2.3.8 Structural Breaks
2.3.9 Data Uncertainty
2.3.10 Forecast Uncertainty Bands
2.4 Theoretical literature review
2.4.1 Regime-Switching Theory and Markov-Switching Models
2.4.2 Vine Copula Theory and Multivariate Dependence Modelling
2.4.3 Empirical Bayesian Credibility Theory
2.4.4 Integrating the Theoretical Streams
2.5 Empirical literature review
2.6 Research Gaps
2.7 Chapter Summary
3 RESEARCH METHODOLOGY
3.1 Introduction
3.2 Research Design
3.3 Research Methodology
3.4 Target Sample
3.5 Data Collection
3.6 Variable Definition and Data Structure
3.7 Regime-Switching Revenue Model
3.8 Multivariate Dependence Modelling
3.9 Bayesian Credibility Adjustment
3.10 Forecasting and Validation
3.11 Data Analysis and Presentation
3.12 Chapter Summary
4 : DATA ANALYSIS AND PRESENTATION
4.1 Introduction
4.2 Data Overview
4.3 Data Exploration
4.3.1 Summary of Tax Revenue Central Tendencies
4.3.2 Correlation Structure of Revenue Heads
4.3.3 Associations with Key Macroeconomic Drivers
4.3.4 Tax Revenue Streams Contribution to Total net Tax
4.3.5 Tax Revenue time series patterns
4.3.6 Revenue Response to Policy and Shock Episodes
4.4 Addressing Research Objectives
4.4.1 Identifying and modelling economic regimes affecting major tax revenue streams using Markov switching techniques
4.4.2 Estimated variance for each regime
4.4.3 Estimated Regime Transition Probabilities
4.4.4 Regime Persistence and Expected Durations
4.4.5 Analysing and quantifying dependency relationships among VAT, PAYE, corporate tax, and customs duties using vine copulas
4.4.6 Overall Dependence Pattern
4.4.7 C Vine Root Choice
4.4.8 Tree 1 Pair Copula Results
4.4.9 Goodness of Fit (Tree 1)
4.4.10 Ordering Signal Cross Check
4.4.11 Developing and evaluating an integrated forecasting model that combines regime switching, dependence structures, and Bayesian credibility for improved predictive performance
4.4.12 Empirical-Bayes and Copula Calibration
4.4.13 Interpretation of Significant EB-Adjusted Coefficients
4.4.14 Out-of-Sample Forecast Accuracy
4.4.15 Graphical Evidence
4.4.16 Loss-difference plots
4.4.17 Summary of Integrated Model Performance
4.5 Hypothesis Testing
4.6 Conclusion
5 :RESEARCH FINDINGS,CONCLUSIONS AND RECOMMENDATIONS
5.1 Introduction
5.2 Summary of findings
5.2.1 Economic Regimes Affecting Zimbabwe’s Major Tax Revenue Streams
5.2.2 Dependency Relationships Among VAT, PAYE, CIT and Customs Duties
5.2.3 Performance of the Integrated Forecasting Framework
5.3 Implications of the study
5.4 Areas for Further Research
5.5 Conclusion
Research Objectives and Core Themes
This study aims to develop and evaluate an integrated econometric forecasting framework for predicting tax revenue collections by the Zimbabwe Revenue Authority (ZIMRA) across its four principal revenue streams—Value Added Tax, Pay As You Earn, Corporate Income Tax, and Customs Duties—by combining Markov-switching regressions, vine copulas, and empirical Bayesian credibility adjustments to handle high volatility, structural instability, and parameter uncertainty.
- Identification of latent economic regimes and state-dependent revenue volatilities using Markov-switching autoregressive models.
- Modelling non-linear and tail dependence among tax heads using regular vine copula decompositions.
- Stabilisation of regime-dependent parameters and noisy administrative data through empirical Bayesian credibility shrinkage.
- Forward simulation and joint revenue distribution generation for credible forecast uncertainty intervals.
- Comparative evaluation of forecasting accuracy against conventional baseline models under macroeconomic shock conditions.
Excerpt from the Book
1.2.5 Advanced tax revenue forecasting under the Integrated Modelling Framework for ZIMRA
The regime-switching models solve structural instability problems by allowing parameters, intercepts and variances to move between different latent states which use Markov processes to model economic policy and currency changes (Hamilton, 1989; Riswanda & Rikumahu, 2024). The risk assessment method known as copula methods together with vine copulas enables researchers to study multiple tax streams through their non-linear and asymmetric relationships while detecting their tail dependencies (Czado & Nagler, 2022; Joe, 1997). The Empirical Bayesian credibility method produces superior estimation results because it combines prior information with observed data to generate better predictions which result in improved performance on new data points (Bühlmann, 1967; Kim, Bezbradica & Crane, 2024). The methods have achieved successful results in financial and macroeconomic fields which experience market instability and major system changes, yet they have not been widely used for tax revenue prediction in Zimbabwe and similar developing nations (Foroni, Merlo & Petrella, 2024; Stephan, 2023).
The three factors which include regime shifts, cross-tax dependence and data uncertainty require multiple methods to determine how they affect revenue results. The framework employs Markov-switching to identify different operational states, vine copulas to analyze non-linear relationships between multiple tax revenue streams and empirical Bayesian credibility for parameter estimation which provides resistance to noise interference, thus matching Zimbabwe’s economic and institutional structure (Czado & Nagler, 2022; Kim, Bezbradica & Crane, 2024). The implementation of the Integration Forecasting Framework has potential to bring operational benefits to ZIMRA as it can enhance prediction accuracy, enabling the organization to handle combined risk events and generate reliable uncertainty ranges which enhance budget reliability, debt control and policy effectiveness. The research study provides a usable model which other developing nations experiencing similar forecasting difficulties can implement (Osinowo, Oyeyemi & David, 2024; Foroni, Merlo & Petrella, 2024).
Chapter Summaries
1 INTRODUCTION: Outlines the research background, statement of the problem, objectives, and significance of modeling tax revenue in Zimbabwe's highly volatile macroeconomic environment.
2 LITERATURE REVIEW: Reviews the theoretical foundation and empirical studies regarding regime-switching models, vine copulas, and empirical Bayesian credibility techniques in fiscal forecasting.
3 RESEARCH METHODOLOGY: Explains the quantitative research design, data collection processes, model specifications, and simulation routines for the integrated forecasting framework.
4 : DATA ANALYSIS AND PRESENTATION: Details empirical estimations including Markov-switching regime classifications, vine copula dependencies, Bayesian adjustments, and out-of-sample forecast evaluations.
5 :RESEARCH FINDINGS,CONCLUSIONS AND RECOMMENDATIONS: Synthesizes the core research findings, discusses policy implications for ZIMRA and the Treasury, and identifies avenues for future research.
Keywords
Tax Revenue Forecasting, Markov-Switching, Vine Copulas, Empirical Bayesian Credibility, ZIMRA, Zimbabwe, Regime Shifts, Asymmetric Dependence, Fiscal Planning, Out-of-Sample Forecasting, Predictive Distribution
Frequently Asked Questions
What is the core focus of this research?
The study focuses on developing an integrated econometric forecasting framework to model and project tax revenues collected by the Zimbabwe Revenue Authority under conditions of high economic volatility and structural change.
What are the central thematic areas examined?
The primary thematic areas encompass macroeconomic instability, non-linear cross-tax dependencies, latent regime transitions, Bayesian parameter shrinkage, and probabilistic forecast intervals.
What is the primary objective of the study?
The primary objective is to enhance the robustness, accuracy, and policy utility of revenue projections for ZIMRA by uniting Markov-switching models, vine copula dependence structures, and empirical Bayesian credibility theory.
What research methodology is applied?
The research uses a quantitative, model-based econometric methodology incorporating two-regime Markov-switching regressions, C-vine copulas on regime-filtered residuals, and empirical Bayes shrinkage estimation executed via Python workflows.
What tax categories are examined in the empirical section?
The empirical analysis focuses on Zimbabwe's four main fiscal revenue sources: Value Added Tax (VAT), Pay As You Earn (PAYE), Corporate Income Tax (CIT), and Customs Duties.
Which keywords characterise the research?
The work is characterised by keywords such as Tax Revenue Forecasting, Markov-Switching, Vine Copulas, Empirical Bayesian Credibility, ZIMRA, and Regime Shifts.
How do economic regimes differ across tax heads in Zimbabwe?
The findings indicate that Corporate Income Tax and VAT exhibit prolonged periods of high volatility and turbulence, whereas PAYE displays stable and balanced regime persistence due to its reliance on formal wages.
Did the integrated framework outperform the baseline model in point forecasts?
While the baseline Markov-switching model achieved slightly lower point forecast error metrics (RMSE and MAPE), the integrated framework demonstrated superior capability in capturing realistic uncertainty bands, joint tail risks, and revenue shocks.
- Quote paper
- Ethaton Mawela (Author), 2026, Modelling Zimra Tax Revenues with Markov-Switching, Vine Copulas and Empirical Bayesian Credibility, Munich, GRIN Verlag, https://www.grin.com/document/1746824