Why it matters
Submitted through the ingestion workflow. Review and refine the category and supporting notes before treating it as a settled entry.
Notes
SVAR (Structural Vector Autoregression) is an econometric tool that tries to separate tangled economic data into the underlying independent surprises — called shocks — that drove it, so you can simulate what happens if one of those surprises hits again. The motivating problem When two things move together in the economy (e.g., interest rates and unemployment), you can't tell from the data alone whether A caused B or B caused A. Both causal stories fit the same numbers. This is the core identification problem. What a VAR does A VAR (vector autoregression) predicts today's values from recent past values across multiple variables simultaneously. It's a correlation machine across time — it tells you what tends to happen next but not why. It captures the dance between variables without knowing who's leading. What the "S" adds The S stands for Structural. An SVAR forces the model to decompose comovements into clean, independent causes — structural shocks. These are the economy's original surprises: a surprise oil price spike, a surprise Fed decision, a surprise productivity boom. Everything else (downstream movements in GDP, unemployment, prices) is just the economy digesting those shocks. Identifying restrictions To go from a VAR to an SVAR, you need extra assumptions called identifying restrictions. These are what allow you to decompose the mixed signal into its pure components. They are never fully testable — two economists can take the same data and disagree about what counts as a "pure monetary shock" based on different assumptions. SVARs aren't magic; they're a disciplined way of being explicit about your assumptions. Analogies - Cocktail: A VAR describes what you taste (sweet, citrusy, bitter). An SVAR reverse-engineers the recipe (30ml gin, 15ml lime). To do that, you need assumptions — like knowing the bartender only uses three ingredients. - Sound engineer: The raw recording with drums, guitar, and vocals mixed together is the VAR. Separating it back into individual tracks is what an SVAR does — but you need assumptions to un-mix the signal. Why it matters Once you have clean shocks, you can run impulse response functions — graphs showing how the economy reacts over time to one pure shock, holding others fixed. Central banks, the IMF, and finance shops use SVARs to answer "what if the Fed does a surprise rate hike tomorrow?" type questions. TL;DR A statistical tool that separates tangled economic data into the independent surprises that drove it, enabling causal what-if simulations.