Calibrated AI-Labor Market Simulation Model

This repository contains the full Python implementation of the calibrated simulation model presented in the paper: "The Heterogeneous Effects of Artificial Intelligence on Labour Markets: A Calibrated Simulation of Skills, Tasks, and Wages." The model provides a quantitative framework for analyzing the dynamic impact of AI on wage inequality, labor displacement, and economic growth. It extends the traditional task-based model by incorporating a novel two-dimensional task space (cognitive complexity and codifiability) and a dynamic AI capability frontier. This code was developed to address the critical need for an empirically grounded, quantitative backbone for theoretical models of AI's economic effects. It is designed to be fully replicable, allowing researchers to verify our findings and extend the model for future work. 2. Key Features The simulation model includes several novel features designed to provide a more realistic analysis of AI's labor market impacts: Empirical Calibration: Core parameters (e.g., elasticity of substitution σ, initial AI frontier θ₀) are calibrated to match stylized facts from the economic literature, such as the high-skill wage premium and aggregate automation risk. Dynamic Simulation: The model evolves over discrete time periods, capturing the progressive expansion of AI capabilities and its cumulative effects on the economy. Two-Dimensional Task Space: Tasks are defined by both cognitive complexity and codifiability, allowing for a more nuanced analysis of which tasks are susceptible to automation. Comparative Advantage Task Allocation: The model allocates tasks to the most efficient agent (AI, low-skill human, or high-skill human) based on their comparative advantage, simulating firm-level cost-minimization. Quantitative Policy Analysis: The framework includes modules for simulating counterfactual policy scenarios (e.g., "AI tax," training subsidies) and quantifying their impact on inequality and growth. 3. Requirements The code is written in Python 3 and relies on several standard scientific computing libraries. To run the simulation, you will need to have the following packages installed: numpy pandas matplotlib seaborn scipy scikit-learn