Towards Causal Market Simulators

TL;DR

This paper introduces TNCM-VAE, a novel deep generative model that incorporates causal reasoning into financial time series generation, enabling more accurate counterfactual analysis and risk assessment.

Contribution

The paper presents a new causal time-series generative model combining VAEs with structural causal models, enforcing causal constraints for realistic counterfactual simulation.

Findings

Superior counterfactual probability estimation with low L1 distances (0.03-0.10).

Enables realistic scenario analysis and stress testing in finance.

Validates approach on synthetic autoregressive models inspired by Ornstein-Uhlenbeck process.

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.