Abstract: I propose identifying structural vector autoregressions using ‘shock-percentile’ restrictions. These restrictions require the realisation of a structural shock in a selected episode to lie in the tail of the shock’s historical distribution, representing the belief that a relatively large shock has occurred. I argue that shock-percentile restrictions are an attractive alternative to imposing numeric bounds on shock magnitudes, which are difficult to credibly elicit. Simulations demonstrate the potential for shock-percentile restrictions to provide identifying information. In two empirical applications, I exploit shock-percentile restrictions to disentangle the relationship between uncertainty and real activity, and to sharpen identification of the macroeconomic effects of US monetary policy.

https://www.rba.gov.au/publications/rdp/2026/2026-01/full.html