Research Title: Identifying Structural Vector Autoregressions via Non-Gaussianity of Potentially Dependent Shocks

Authors: Markku Lanne, Liu Keyan, Jani Luoto

Journal: The Econometrics Journal, published online February 2026

Summary:

Structural Vector Autoregression (SVAR) models have long been important tools in macroeconomic analysis for economic shocks and dynamic impacts. However, effectively identifying structural shocks without traditional zero restrictions or exogenous instrumental variables remains core problem in econometrics.

Key Findings:

1. Using higher-order moment perspective, study systematically explores structuralshocks' skewness and kurtosis roles in identification

2. Paper proves: under condition of no structural shock co-skewness, skewed structural shocks can be identified; under condition of no excess co-kurtosis, shocks with nonzero excess kurtosis can be identified

3. Under this framework, remaining shocks may exhibit set identification features

4. In skewness-based identification scenario, model allows structural shocks to have correlated conditional heteroskedasticity, significantly expanding applicability of non-Gaussian identification methods

Methodology: Builds Bayesian SVAR model with skewed t-distribution error terms and correlated stochastic volatility to better capture non-Gaussian data features

Empirical Application: Applied to US monetary policy empirical analysis, validatingtheoretical results' operationality and explanatory power in real macroeconomic data

Contributions:

1. Completes and expands higher-order moment-based non-Gaussian SVAR identification theory

2. Allows identification under correlated heteroskedasticity shocks, expanding theoretical boundaries

3. Combining theoretical identification results with Bayesian estimation enhances model's application value in macroeconomic empirical research

Journal: The Econometrics Journal(B+ journal under Wuhan University Economics and Management School grading scheme)

Link: //doi.org/10.1093/ectj/utag002