SpinSVAR: Estimating Structural Vector Autoregression Assuming Sparse Input

TL;DR

SpinSVAR is a new method for estimating structural vector autoregressions that assumes sparse, Laplacian-distributed inputs, providing theoretical guarantees and superior performance on synthetic and real financial data.

Contribution

It introduces a sparse input assumption with Laplacian noise for SVAR estimation, along with a scalable, GPU-accelerated algorithm and theoretical consistency guarantees.

Findings

Outperforms existing methods in accuracy and speed on synthetic data.

Effectively clusters stocks by sectors and detects key shocks in financial data.

Demonstrates the practical viability of sparse input assumptions in real-world applications.

Abstract

We introduce SpinSVAR, a novel method for estimating a structural vector autoregression (SVAR) from time-series data under sparse input assumption. Unlike prior approaches using Gaussian noise, we model the input as independent Laplacian variables, enforcing sparsity and yielding a maximum likelihood estimator (MLE) based on least absolute error regression. We provide theoretical consistency guarantees for the MLE under mild assumptions. SpinSVAR is efficient: it can leverage GPU acceleration to scale to thousands of nodes. On synthetic data with Laplacian or Bernoulli-uniform inputs, SpinSVAR outperforms state-of-the-art methods in accuracy and runtime. When applied to S&P 500 data, it clusters stocks by sectors and identifies significant structural shocks linked to major price movements, demonstrating the viability of our sparse input assumption.