svar shock identification

This paper addresses the challenges of shock identification, excessive constraints, and computational intractability in high-dimensional structural vector autoregressions (SVARs). Methodologically, it introduces a unified framework for jointly imposing linear inequality constraints on both the shock space and the impact space—specifically modeling sign and magnitude restrictions simultaneously on structural shocks, contemporaneous impulse responses, and their element-wise products. It develops a scalable Bayesian MCMC sampling algorithm that overcomes computational bottlenecks in high-dimensional settings (up to 30 variables) under multiple inequality constraints. The key contributions are: (i) precise identification and differentiation of structural shocks, successfully isolating five distinct macroeconomic shocks; and (ii) empirical evidence identifying financial shocks as the dominant driver of business cycle fluctuations. The proposed framework substantially enhances inference accuracy and structural interpretability in SVAR analysis.

This paper addresses the challenge of identifying specific structural shocks in structural vector autoregressive (SVAR) models using heteroskedasticity alone—without conventional sign or exclusion restrictions. Within a Bayesian framework, we propose a non-centered stochastic volatility approach that dispenses with such auxiliary constraints. Theoretically, we derive necessary and sufficient conditions for partial and global identification of structural parameters. Methodologically, we develop a heteroskedasticity-based statistical identification diagnostic and introduce a shrinkage prior centered at homoskedasticity, ensuring identification is fully data-driven. Empirically, applying the method to a U.S. fiscal structural model, we achieve partial identification of structural shocks without imposing additional identifying restrictions, thereby substantially enhancing estimation robustness and economic interpretability.

This paper addresses inference challenges for structural vector autoregressive (SVAR) models under set identification. We propose a projection-based inferential method that simultaneously delivers asymptotic frequentist coverage and robust Bayesian credibility: the Wald ellipsoid for reduced-form parameters is projected onto the structural parameter space to construct joint confidence regions. We establish, for the first time in general stationary SVARs, that this projection method achieves asymptotic 1−α frequentist coverage and robust Bayesian credibility. Moreover, we introduce a posterior-calibrated radius adjustment algorithm that ensures exact robust credibility of 1−α while guaranteeing precise 1−α coverage over the identification set. Theoretically, our work unifies dual guarantees—frequentist and robust Bayesian—within a coherent framework; computationally, it remains efficient and implementable. Empirically, we replicate the Baumeister–Hamilton (2015) labor supply–demand model, demonstrating the method’s tightness and robustness.

This paper addresses the statistical identifiability of structural thresholds and smooth-transition VAR models under shock independence and at most one Gaussian shock. Methodologically, it introduces the first extension of independent non-Gaussian identification to a nonlinear SVAR framework featuring time-varying structural matrices, thereby transcending conventional zero-restriction reliance. Identification is achieved globally solely through shock non-Gaussianity, resolving shock-label ambiguity across regimes. The approach integrates independent component analysis, non-Gaussianity testing, and logistic smooth-transition mechanisms, with implementation supported by the R package *sstvars*. Empirically, the framework reveals that climate policy uncertainty shocks exert persistent inflationary effects—significantly amplified during periods of high economic policy uncertainty—and robustly suppress output.

This paper investigates whether linear impulse response methods—such as vector autoregressions (VARs) and local projections—retain causal interpretability when the true data-generating process is nonlinear. Using theoretical derivation and sensitivity analysis, we establish, for the first time, that standard linear estimators robustly identify a *weighted average* of causal effects—not point-identified effects—under nonlinearity, whereas identification strategies relying on heteroskedasticity or non-Gaussianity fail. We then propose a novel theoretical framework based on weighted regression to identify marginal treatment effects. Our analysis precisely characterizes the robustness boundary of linear estimators in nonlinear macroeconomic models, thereby providing a formal foundation for causal interpretation of empirical macroeconomic shock effects. By bridging linear estimation practice with nonlinear structural foundations, this work extends the scope of causal inference to broader nonlinear settings while preserving tractability and interpretability.

This study addresses the challenges of identifying country-specific macroeconomic shocks and their international transmission in multi-country systems—namely, the curse of dimensionality, heavy computational burden, and proliferating identification restrictions—by proposing a Bayesian Structural Matrix Autoregressive (BSMAR) framework. The approach exploits the natural matrix structure of international macroeconomic data to disentangle cross-variable and cross-country dependencies, enabling parsimonious modeling of large systems. It innovatively integrates conventional SVAR identification strategies with a novel method for identifying contemporaneous international spillovers, coherently accommodating zero restrictions, sign restrictions, and ordering constraints. Empirical analysis using quarterly data from 15 economies reveals substantial heterogeneity in cross-border shock transmission and demonstrates that demand shocks play a more prominent role than supply shocks in generating international spillovers.

This study addresses the limitations of traditional structural vector autoregressive (SVAR) models in capturing nonlinear, non-additive contemporaneous relationships and in achieving sufficient identification for independent innovation analysis. To overcome these challenges, the authors propose a fully nonlinear SVAR framework that leverages exogenous-variable-induced shifts in the conditional distribution of structural shocks. By combining exponential family assumptions with contrastive learning and employing feedforward neural networks to estimate the nonlinear system, the approach reduces identification ambiguity from arbitrary invertible transformations to mere permutation and sign indeterminacies. This substantially enhances structural identification accuracy. Empirical results reveal that U.S. industrial production responds to real oil price shocks with modest asymmetries in both sign and economic regime. The proposed methodology is implemented in the open-source R package iiasvar.

This study addresses the sensitivity of traditional structural vector autoregressive (SVAR) models to ad hoc variable selection by proposing a Bayesian framework that incorporates an algorithm-driven variable selection mechanism. The approach integrates recursive identification with a Bayesian SVAR, an anchor-free joint proxy model, and a multi-instrument strategy, enabling automatic construction and optimization of the information set via out-of-sample criteria. This method preserves the largest feasible system while enhancing model robustness and interpretability. Empirical results reveal that housing production—not household credit—is the primary driver of output expansion. Furthermore, in monetary policy transmission, the credit spread channel is markedly amplified, with corporate default risk emerging as a critical margin.

This study addresses the computational inefficiencies and high autocorrelation commonly encountered in traditional Bayesian inference for high-dimensional structural vector autoregressive (SVAR) models with symbolic constraints. The authors propose a reparameterization-based Hamiltonian Monte Carlo (HMC) algorithm that unifies inequality identification constraints—such as shape, ordering, and elasticity bounds—with zero restrictions into a continuously differentiable mapping. This approach enables, for the first time, differentiable handling of complex mixed constraints within SVAR estimation. Empirical results demonstrate that the method substantially improves posterior sampling efficiency, yielding lower autocorrelation, higher effective sample sizes, and reduced computation time. Consequently, the proposed framework enhances both the stability and scalability of Bayesian inference in high-dimensional SVAR models.