Post

🧵: What does monetary policy do? Our best evidence comes from quality empirical monetary policy shocks (EMPS) using high-frequency and narrative methods. … but what *are* these shocks?

Philip Barrett @phibar.bsky.social and I address this question in a new working paper. In it, we propose a method to extract the news that EMPS contain about future policy, and apply the method to many shocks from the literature. There is a lot to learn!

Background: EMPS are used to evaluate our theories of monetary policy. But this cannot really be done without knowing what the EMPS *are*. EMPS are not just textbook target rate surprises. They also contain “news shocks” about future policy.

Important: surprise shocks have different effects from news shocks at every news horizon. Here’s how they affect prices in the New Keynesian model:

So, if we want to use EMPS to evaluate our theories, we absolutely need to know how they depend on surprise vs news over many horizons. We call this the *Term Structure of Monetary Policy News*. Fortunately: there is a way to estimate this term structure!

Estimation has 4 stages: 1. Instrument for endogenous variables using macro IVs (Barnichon-Mesters) 2. Estimate policy rule by IV 3. Whiten the policy residual to find the innovation 4. Regress on past EMPS (local projection) The coefficients from the final stage = term structure

This is an unbiased estimator of the term structure! We also derive a smooth-local-projection variant. Convenient: both can be written as a single-stage estimator with analytical standard errors (no bootstrap!)

Application #1: We estimate the term structures of popular EMPS in the literature. E.g. here is the term structure for Swanson’s forward guidance shock. Nearly all the information is news about future policy.

Finding: *all shocks* are mostly news. Even the most surprise-like shocks are <50% due to immediate policy changes. Here’s the Aruoba-Drechsel narrative shock:

Unwelcome news if you wanted to use EMPS to study textbook policy surprises? Fortunately, we offer a way to do that (and more!) Using our method, it is possible to construct a *synthetic* monetary policy shock approximating any desired term structure of policy news!

Application #2: We estimate the macro effects of synthetic MPS. - The synthetic surprise is contractionary, but has no effect on prices. - Instead, deflationary effects of (some) EMPS are due to their long-run news components.

Bonus: IV estimation of the Taylor rule works very well. Reliable, robust. If you are interested in this method, you can find the macro shocks we use here: github.com/jonathanjada...