Simulation & Synthetic Data
Agent-based modeling, synthetic data generation, and computational methods for economic simulation • 24 papers
Agent-Based Modeling
Model complex economic systems through heterogeneous interacting agents
Agent-Based Computational Economics: A Constructive Approach to Economic Theory
Foundational survey establishing ACE methodology. Defines agent-based computational economics as the computational study of economies modeled as evolving systems of autonomous interacting agents. Published in Handbook of Computational Economics Vol. 2.
Agent-Based Modeling in Economics and Finance: Past, Present, and Future
Definitive 90-page survey in Journal of Economic Literature reviewing 30 years of ABM in economics. Covers finance, industrial organization, macroeconomics, and policy applications. Essential reading for the field.
Artificial Economic Life: A Simple Model of a Stockmarket
The Santa Fe Artificial Stock Market—foundational heterogeneous agent financial model. Shows how bubbles and crashes emerge from adaptive learning traders. Seminal work in agent-based finance.
The Economy Needs Agent-Based Modelling
Nature commentary arguing for ABM in economic policy. Critiques DSGE models' failure to predict 2008 crisis; calls for bottom-up agent-based approaches to model systemic risk.
Leverage Causes Fat Tails and Clustered Volatility
Agent-based explanation of financial market fat tails and volatility clustering. Shows leveraged agents create price fluctuations 10x larger than fundamentals. Published in Quantitative Finance.
Mesa: An Agent-Based Modeling Framework
Introduces Mesa, the leading open-source Python ABM framework. Describes modular architecture with spatial grids, agent schedulers, and data collection. Published in JOSS.
Econ-ARK and HARK: Open-Source Tools for Computational Economics
Describes Econ-ARK ecosystem and HARK toolkit for heterogeneous agent modeling. NumFOCUS-sponsored project implementing Aiyagari, buffer-stock, and lifecycle models.
A Standard Protocol for Describing Individual-Based and Agent-Based Models
The ODD (Overview, Design concepts, Details) protocol for describing ABMs. Standard framework for model documentation adopted across ecology and social simulation.
Synthetic Data Generation & Privacy
Generate privacy-preserving synthetic datasets that maintain statistical properties
Modeling Tabular Data using Conditional GAN
Introduces CTGAN using mode-specific normalization and conditional generator for tabular data synthesis. Handles mixed discrete/continuous columns. Published at NeurIPS.
DataSynthesizer: Privacy-Preserving Synthetic Datasets
Introduces DataSynthesizer for generating synthetic data with differential privacy using Bayesian networks. Three modes: random, independent attributes, and correlated attributes.
synthpop: Bespoke Creation of Synthetic Data in R
Introduces synthpop R package using CART-based synthesis. Produces realistic synthetic data preserving statistical relationships. Published in Journal of Statistical Software.
The Synthetic Data Vault
Introduces SDV system for synthesizing relational databases while preserving referential integrity. Uses copulas and deep learning for multi-table synthesis.
PATE-GAN: Generating Synthetic Data with Differential Privacy Guarantees
Combines Private Aggregation of Teacher Ensembles (PATE) with GANs for differentially private synthetic data. Provides formal privacy guarantees with strong utility.
Anonymeter: A Python Library for Privacy Risk Assessment
Framework for GDPR-aligned privacy risk quantification. Measures singling out, linkability, and inference risks for synthetic datasets.
Mechanism Design & Reinforcement Learning
Design optimal economic mechanisms using deep reinforcement learning
Optimal Auctions through Deep Learning
Introduces RegretNet—neural network architecture that learns approximately revenue-optimal auctions. First general-purpose approach to automated mechanism design. Published at ICML, later CACM 2020.
The AI Economist: Taxation Policy Design via Two-Level Deep Reinforcement Learning
Two-level RL for optimal taxation: agents optimize labor/trade, social planner optimizes tax policy. Finds policies outperforming baselines by 16% on equality-productivity. Published in Science Advances.
Multi-Agent Reinforcement Learning in Sequential Social Dilemmas
Foundational work on multi-agent RL in games with mixed cooperative/competitive incentives. Introduces sequential social dilemmas; shows cooperation emergence depends on environment structure.
Deep Reinforcement Learning for Optimal Execution
Applies deep RL to optimal order execution problem. Learns execution policies that outperform TWAP/VWAP benchmarks on real market data.
Discovering Algorithmic Pricing in Practice: A Framework for Learning-Based Collusion
Studies emergence of tacit collusion in RL-based pricing algorithms. Shows Q-learning agents can learn supra-competitive prices without explicit coordination.
Market Microstructure Simulation
Simulate limit order book markets and financial exchange dynamics
ABIDES: Towards High-Fidelity Multi-Agent Market Simulation
Introduces ABIDES (Agent-Based Interactive Discrete Event Simulation) from JPMorgan. NASDAQ-like exchange with ZI, momentum, and market maker agents. Published at AAMAS.
Get Real: Realism Metrics for Robust Limit Order Book Market Simulations
Establishes stylized facts benchmarks for evaluating LOB simulation realism. Proposes metrics for fat tails, volatility clustering, and bid-ask spread dynamics.
Internet Advertising and the Generalized Second-Price Auction
Analyzes the Generalized Second-Price (GSP) auction used by Google and Yahoo for sponsored search. Shows GSP has efficient Nash equilibria equivalent to VCG outcomes. Published in AER.
Limit Order Book Simulation Methods: A Review
Comprehensive review of LOB simulation methods: order flow models, agent-based approaches, and deep learning. Covers validation metrics and calibration techniques.
AuctionGym: Simulating Online Advertising Auctions
Amazon's ad auction simulator for training RL bidding agents. Supports first/second-price auctions with realistic features. Best Paper at AdKDD 2022.