Towards Causal Market Simulators
Dennis Thumm, Luis Ontaneda Mijares
TL;DR
The paper tackles the lack of causal reasoning in market data generators by introducing TNCM-VAE, a time-series variational autoencoder augmented with structural causal modeling. It enforces causal constraints via a DAG-structured decoder and trains with a causal Wasserstein objective, enabling counterfactual generation in financial time series. Validation on OU-inspired autoregressive data shows accurate counterfactual probability estimates (L1 0.03–0.10), highlighting improvements in stress testing, scenario analysis, and backtesting. This approach offers a principled framework for generating plausible, causally consistent market trajectories useful for risk management and decision support.
Abstract
Market generators using deep generative models have shown promise for synthetic financial data generation, but existing approaches lack causal reasoning capabilities essential for counterfactual analysis and risk assessment. We propose a Time-series Neural Causal Model VAE (TNCM-VAE) that combines variational autoencoders with structural causal models to generate counterfactual financial time series while preserving both temporal dependencies and causal relationships. Our approach enforces causal constraints through directed acyclic graphs in the decoder architecture and employs the causal Wasserstein distance for training. We validate our method on synthetic autoregressive models inspired by the Ornstein-Uhlenbeck process, demonstrating superior performance in counterfactual probability estimation with L1 distances as low as 0.03-0.10 compared to ground truth. The model enables financial stress testing, scenario analysis, and enhanced backtesting by generating plausible counterfactual market trajectories that respect underlying causal mechanisms.