Evidence and data

The identifying restriction is the research design. Report the restriction, horizon, data vintage, and sensitivity checks.

Calibration or measurement

Ordering, sign restrictions, instruments, and shock normalization determine the structural impulse response.

Boundaries

• Different admissible restrictions can tell different stories.
• Weak instruments damage proxy SVARs.
• Historical decompositions inherit identification risk.

Use guidance

When sufficient

Identified impulse responses to a small number of structural shocks when the identification scheme, whether recursive ordering, sign restrictions, long-run restrictions (Blanchard-Quah 1989 AER), or an external instrument, is credible and can be defended. The standard tool for estimating the effect of a monetary-policy shock, a government-spending shock, or an oil-supply disruption when theory restricts the impact or frequency-domain behavior of the shock.

When sketch only

Do not use when the identifying restriction is itself the contested question. If the debate is whether central banks reacted to output or output reacted to central banks, a recursive SVAR that resolves this by assumption is not a structural test; it is a labeling exercise. Sign-restricted SVARs can also be under-identified, with a wide posterior set of admissible impulse responses.

When to switch

Switch to an external-instrument SVAR (Mertens-Ravn 2013 AER; Stock-Watson 2018 JEEA proxy-SVAR) when a narrative or high-frequency surprise measure provides a stronger instrument than within-sample restrictions. Switch to an estimated DSGE (dsge:nk) when the structural interpretation of each shock needs to be pinned to a specific friction rather than a timing or sign assumption.

Falsification signal

Impulse responses that flip sign or change magnitude substantially across plausible but distinct identification schemes on the same dataset indicate the result is not robust to identification uncertainty. A monetary shock that appears contractionary under recursive ordering but expansionary under sign restrictions signals that the data do not sharply identify the shock, regardless of which label is assigned.