cyrusCMM/PAYE-Tax-Simulation-Kenya

Simulation of alternative PAYE rate structures and revenue outcomes using administrative payroll data. # PAYE Tax Simulation – Kenya # Simulation of PAYE tax policy alternatives using administrative payroll data to evaluate revenue–equity trade-offs. This project builds a grid of alternative PAYE marginal tax rate structures and evaluates their impact on government revenue and tax progressivity using the Kakwani index. The repository contains: - A full PAYE microsimulation notebook - A grid of simulated tax systems - Revenue and progressivity metrics - Visualization dashboards of the policy space and Pareto frontier The analysis constructs a structured grid of marginal tax rate combinations and evaluates each tax system in terms of both **revenue performance** and **tax progressivity**. The resulting policy space is analyzed to identify the **Pareto frontier of efficient tax systems**, highlighting the trade-off between revenue generation and equity. # Project Objective The objective of this project is to analyze how alternative marginal PAYE tax rate structures affect: - Total PAYE revenue - Revenue changes relative to the baseline tax system - The distribution of tax payments across the income distribution - The progressivity of the tax system Using taxpayer-level administrative employment income data, the simulation evaluates thousands of alternative tax systems and identifies **efficient PAYE structures that improve revenue and/or progressivity**. # PAYE Structure Used in the Model The model begins with the current Kenyan PAYE system consisting of **five marginal tax bands**. The following parameters are held constant: - **Band 1 marginal rate (r1): 10%** - **Personal relief: KES 28,800** - **Statutory income thresholds** The remaining marginal rates are varied within a structured policy grid. | Rate | Simulated Range | |-----|----------------| | r2 | 25% – 30% | | r3 | 30% – 32.5% | | r4 | 32.5% – 35% | | r5 | 35% – 40% | All simulated tax systems satisfy the progressive ordering condition: r1 ≤ r2 ≤ r3 ≤ r4 ≤ r5 # Data Preparation The simulati …