Search papers, labs, and topics across Lattice.
This paper introduces CIR-ACTIVA, an amortized model for estimating distributional causal effects in multivariate Cox-Ingersoll-Ross (CIR) processes, addressing the limitations of traditional CIR models that only capture correlated co-movement. By framing trajectories as time-stamped observations, CIR-ACTIVA predicts multi-horizon shock responses without the need for retraining, and it is validated against synthetic ground truth and real CDS data. The results show that CIR-ACTIVA outperforms existing causal-inference baselines in terms of causal selectivity and horizon-resolved calibration, particularly excelling at short horizons, thus enabling new types of what-if analyses in financial forecasting.
CIR-ACTIVA reveals that causal forecasting can outperform traditional methods, allowing for precise multi-horizon predictions in financial markets without retraining.
Mean-reverting dynamics are pervasive in finance, and the Cox--Ingersoll--Ross (CIR) process is a standard model for the time series they produce, from short rates to credit default swap (CDS) spreads. Yet CIR models capture only \emph{correlated} co-movement, not \emph{causal} influence between series, so they cannot answer the system's response when one series is externally shocked, which observational conditionals confound with historical co-movement. We make two contributions. First, an amortized model for distributional causal effect estimation that frames trajectories as time-stamped observations and predicts the calibrated multi-horizon shock response without retraining per scenario. Second, a causal multivariate CIR data-generating process that supplies the paired observational and interventional ground truth that real markets cannot. We instantiate and calibrate the framework on CDS spreads as a testbed. CIR-ACTIVA's validity is established on synthetic ground truth, independent of how well the simulator matches reality, while practical grounding is assessed by backtesting the generated traces against real CDS data. Against observational and amortized causal-inference baselines, CIR-ACTIVA leads on both causal selectivity in the joint distribution and horizon-resolved calibration, retaining its selectivity once the interventional law varies over the horizon, with gains concentrating at short horizons. This opens up a class of what-if queries on coupled spread systems, CDS stress testing among them, that observational forecasters cannot answer.
