Identifying three types of monetary policy surprises in the U.S.
This paper by Marek Jarociński and Peter Karadi addresses a fundamental challenge in monetary policy identification: high-frequency interest rate surprises around Federal Reserve announcements are contaminated by two confounding factors beyond pure monetary policy shocks—central bank information (CBI) effects and Fed response to news (FRN) effects. Jarociński and Karadi develop a methodology to simultaneously identify all three shock types and assess their relative importance. More specifically, they decompose monetary policy surprises (MPS) into three distinct components:
- Monetary policy (MP) shock: A temporary deviation from the systematic policy rule, exogenous to the state of the economy. Theoretically, it causes stock prices to fall immediately (higher discount rate + lower expected dividends) and economic activity to decline.
- Central bank information (CBI) shock: Arises when the Fed reveals information about the economy that differs from market assessments. A favorable CBI shock (Fed signals stronger economy) causes both interest rates and stock prices to rise simultaneously—the classic “wrong-signed” response that violates textbook predictions.
- Fed response to news (FRN) shock: Arises when markets systematically underestimate the Fed’s aggressiveness in responding to publicly available economic news. When favorable news arrives and the Fed responds more hawkishly than expected, stock prices fall (tighter policy) but economic activity subsequently strengthens (due to the underlying good news).
The paper combines three sources of identifying information:
A key methodological innovation is exploiting heteroskedasticity between FOMC and non-FOMC events. The authors show that:
- FRN shocks are more prevalent at FOMC announcements (where Bauer-Swanson regressions explain ~19% of surprises)
- CBI shocks are more prevalent at non-FOMC events (Fed chair speeches, minutes releases)
The paper extends the event sample from 358 FOMC announcements to over 1,600 Fed communication events (1988–2024), including Fed chair speeches, congressional testimonies, and FOMC minutes releases. This expansion is crucial because the R-squared of Bauer-Swanson predictive regressions drops from 19% (FOMC only) to 6% (combined sample), indicating that FRN effects are much weaker in non-FOMC events.
Findings
Central bank information effects are robust
Even when carefully controlling for FRN effects, CBI shocks remain a significant feature of the data. Favorable CBI shocks generate:
- Immediate stock price appreciation
- Higher breakeven inflation rates
- Lower corporate bond spreads and VIX
- A temporary boom in economic activity that monetary policy offsets within 1–2 years
Fed response to news effects are limited at monthly frequencies
While FRN shocks show significant contractionary effects in daily local projections (consistent with markets learning about tighter-than-expected systematic policy), their impact in monthly VARs is marginal. The authors attribute this to the two components of FRN shocks—stimulative fresh public news and contractionary policy surprise—largely offsetting each other at monthly horizons.
Purified MP shocks match theoretical predictions
Monetary policy shocks cleaned of both CBI and FRN contamination generate responses consistent with standard theory:
- Stock prices decline
- Financial conditions deteriorate (excess bond premium rises)
- Real GDP contracts
- Price level falls
These findings contrast with “naive” approaches using raw interest rate surprises, which suffer from attenuation bias because both CBI and FRN effects create positive correlation between rate surprises and economic activity.
Disentangling Monetary Policy, Central Bank Information, and Fed Response to News Shocks
Authors: Marek Jarociński, Peter Karadi
From: European Central Bank
Disentangling policy from information in real time in Switzerland
The paper by David Borner addresses how to accurately classify monetary policy announcements into pure monetary policy versus central bank information motives, and how these different types of surprises affect Swiss financial markets. He introduces the “Trading Markov Sign Restriction” (TMSR)—a method that explicitly models high-frequency trading dynamics and expectation adjustment phases within announcement windows. This refinement of the “Poor Man’s Sign Restriction” (PMSR) originally proposed by Jarociński & Karadi (2020) (the authors of the above paper) shows how their earlier work has inspired methodological extensions. The TMSR explicitly models intra-window trading dynamics within the 30-minute event window around Swiss National Bank (SNB) announcements, recognizing that simple endpoint-to-endpoint comparisons can misclassify motives. The methodology proceeds in three steps:
1. Markov switching model for phase detection
Using a Markov switching model on 30-second bucket volatilities of interest rate futures trading volumes, Borner identifies three systematic expectation adjustment phases around SNB announcements:
- Pre-announcement betting phase (t−10 to t0): minimal rate movement and lowest trading volume volatility
- High-intensity positioning phase (t0 to t+5): abrupt rate movements and highest trading volume volatility
- Post-announcement tuning phase (t+5 to t+20): moderate rate movements and considerable but declining volatility
2. Phase-weighted trading direction classification
For each phase, the paper runs time-trend regressions on both interest rate futures and stock market prices to determine trading directions. Sign restrictions classify motives: negative co-movement (rates up, stocks down) indicates pure monetary policy; positive co-movement indicates central bank information.
3. Trading Markov (TM) weights
Crucially, the pre-announcement phase receives zero weight (the central bank is silent), while post-announcement phases are weighted by their duration-adjusted trading volumes to reflect expectation adjustment intensity.
Findings
The TMSR classifies approximately 60% of non-zero SNB monetary surprises as pure monetary policy motives—10 percentage points more than the PMSR. The methods disagree for 50% of all non-zero surprises, particularly for larger pre-financial-crisis shocks.
The critical finding is that pure monetary policy and central bank information surprises affect interest rates and bonds similarly, but exhibit statistically different effects only for exchange rates and stock markets. This contrasts with U.S. and Euro Area studies where the distinction matters more broadly across asset classes.
Central bank information and pure monetary policy surprises in Switzerland
Authors: David Borner
From: Swiss National Bank
