Identifying Monetary Policy Shocks via Institutional Support Restrictions
Abstract
This paper studies the identification of monetary policy shocks from high-frequency announcement surprises. I argue that institutional knowledge of when policy margins are inactive provides naturally arising identifying restrictions. In a factor model, I represent shocks as factors with zero regime-specific variances when inactive, creating a boundary case of identification through heteroskedasticity. For a fixed shock support pattern, I establish local identification of the corresponding covariance model with time-invariant loading vectors normalised to unit length whenever its covariance Jacobian has full rank. Subject to sign and label conventions, this identifies the model parameters.
I revisit Swanson (2021)'s identification strategy using a macroeconomic model that reveals why his zero-loading restrictions need not hold, while also motivating support changes through discontinuities in policy rules and state dependence in the effects of quantitative easing, depending on whether the lower bound on the policy rate binds or is slack. For a specification based on his setting, I show that identification is still possible while allowing the transmission of this shock alone to differ across these two regimes and that empirical estimates provide statistically significant evidence of state dependence, particularly in the responses of medium- and long-term yields.