Working Paper: CEPR ID: DP13853
Authors: Silvia Miranda-Agrippino; Giovanni Ricco
Abstract: This paper discusses the conditions for identification in SVAR-IVs when only the shock of interest or a subset of the structural shocks can be recovered as a linear combination of the VAR residuals. This condition of partial invertibility is very general, often of empirical relevance, and less stringent than the standard full invertibility that is routinely assumed in the SVAR literature. We show that, under partial invertibility, the dynamic responses can be correctly recovered using an external instrument even when this correlates with leads and lags of other invertible shocks. We call this a limited lead-lag exogeneity condition. We evaluate our results in a simulated environment, and provide an empirical application to the case of monetary policy shocks.
Keywords: Identification with external instruments; Structural VAR; Invertibility; Monetary policy shocks
JEL Codes: C32; C36; E30; E52
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| partial invertibility (C20) | identification of structural shocks (C22) |
| external instruments (O36) | identification of structural shocks (C22) |
| limited lead-lag exogeneity condition (C51) | identification of dynamic responses (C32) |
| weaker conditions (C62) | broader applicability of SVAR methodologies (C32) |
| partial invertibility (C20) | identification of causal effects despite misspecifications (C50) |
| impact effects of shocks (F69) | accurate estimation despite biases in IRFs (C51) |
| contamination of instruments vs model misspecifications (C50) | stable impact responses (C62) |