Why are supposedly exogenous monetary policy surprises, measured by changes in short-term financial contracts within short windows around FOMC announcements, partially predicted by pre-meeting economic and financial information? We propose a new explanation: the Federal Reserve targets economic variables by responding primarily to financial conditions while adopting a ‘wait-and-see’ approach to recent economic data. When markets expect the Fed to target economic variables directly, this creates the predictable component of policy surprises. Using daily-frequency economic and financial data from 2000–2019, we find three pieces of supporting evidence: First, the previously documented strong predictors are reflected in financial markets and not in the Fed’s private information. Second, controlling for financial conditions, recent real economic surprises negatively predict policy surprises, which supports the ‘wait-and-see’ hypothesis over a more aggressive response to economic news (Bauer and Swanson, 2023b). Third, financial conditions alone predict policy surprises as effectively as all other documented predictors combined.
TL;DR: High-frequency Fed policy surprises have been partially predictable from pre-FOMC data for three decades — a puzzle for the efficient market hypothesis. Chen (2026, Journal of Macroeconomics) resolves it: the Fed targets economic outcomes by responding primarily to financial conditions while adopting a wait-and-see stance on recent economic data. Markets take the dual mandate literally and miss this channel. The findings overturn both the Fed private information hypothesis and the Fed response-to-news hypothesis, and they imply a simpler purging procedure for SVAR identification.
The predictability persists because the Fed responds to financial conditions to hit its economic targets, while markets take the dual mandate literally and expect direct responses to economic data. This gap is structural, not a learning failure — which is why decades of observation have not closed it.
The puzzle itself is well-established: Bauer and Swanson document that a handful of pre-announcement variables predict a non-trivial share of high-frequency policy surprises , and Cieslak shows markets systematically underestimate the Fed’s response to economic fluctuations, especially in downturns . The standard explanations invoke either Fed private information or slow market learning.
Both explanations struggle with persistence. Chen (2026) resolves this by showing the Fed primarily reacts to financial conditions — which already embed market expectations and forward-looking information — while adopting a “wait-and-see” stance on recent economic data releases . Markets, taking Chair Powell’s “we don’t target financial conditions” literally, miss this channel entirely.
Three market blind spots generate the predictability:
Evidence snapshot: Controlling for a daily financial stress index and Treasury skewness alone reduces the predictive R² of the full Bauer-Swanson predictor set from ~12% to under 1% for scheduled FOMC meetings.
| Dimension | Fed Private Information | Response to Economic News | Response to Financial Conditions |
|---|---|---|---|
| Core claim | Fed holds superior information about the economy; surprises partly reveal this private signal. | Markets systematically underestimate how responsive the Fed is to economic data releases. | Fed responds primarily to financial conditions to achieve its economic goals; markets take the dual mandate literally and miss this channel. |
| Key references | Romer & Romer (2000) , Nakamura & Steinsson (2018) , Miranda-Agrippino & Ricco (2021) | Cieslak (2018) , Bauer & Swanson (2023b) , Schmeling et al. (2022) | Caldara & Herbst (2019) , Brunnermeier et al. (2021) , Caballero et al. (2024) , Chen (2026) |
| Testable prediction | Predictors of surprises contain information not already in market prices. | Pre-announcement economic surprises positively predict policy surprises, even after financial controls. | Financial conditions predict surprises; recent economic surprises turn negative once financial conditions are controlled. |
| Empirical verdict | Rejected. Greenbook forecasts lose predictive power after controlling for public info ; Bauer-Swanson predictors already explain 57% of pre-FOMC FSI variation . | Not supported once financial conditions enter. Real-activity surprises within 14 days flip to a negative coefficient , opposite to the news-response sign. | Supported. FSI + Treasury skewness alone drive R² from ~12% to <1% relative to the full Bauer-Swanson set ; sign on FSI is consistently dovish-to-stress. |
| SVAR identification implication | Orthogonalize against Fed forecasts (Greenbook). | Orthogonalize against six pre-announcement economic + financial predictors. | Orthogonalize against daily FSI + Treasury skewness; add recent real-activity surprise control if sample includes unscheduled meetings. |
| Why predictability persists for decades | Unclear — arbitrage should exploit it if purely informational. | Unclear — markets should eventually learn the true reaction parameter. | Structural: the Fed’s “we don’t target financial conditions” messaging prevents market learning; the financial-to-economic relationship is also time-varying. |
| Named concept | Fed information effect | Fed response-to-news effect | Wait-and-see channel · Financial-conditions-sufficiency (Chen 2026 ) |
No — the pre-announcement variables that predict policy surprises are already priced into daily financial conditions, so they cannot be the Fed’s private information.
The “Fed information effect” originates with Romer and Romer, who found Fed forecasts outperform commercial forecasts for inflation , and was sharpened by Nakamura and Steinsson, who interpret the positive co-movement of surprises and private GDP forecasts as evidence the Fed reveals information . Miranda-Agrippino and Ricco build on this by orthogonalizing surprises against Greenbook forecasts .
The evidence has eroded this view. Bauer and Swanson show Greenbook forecasts lose predictive power after controlling for public information , and Lunsford finds the information effect holds in the early 2000s but not afterward . Cieslak and Schrimpf decompose surprises and find information shocks play a minor role at FOMC announcements .
Chen (2026) provides direct evidence against private information: the six strong predictors in Bauer and Swanson explain 57% of variation in the OFR Financial Stress Index the day before FOMC meetings, meaning their information content is already embedded in market prices . The Fed and the market see the same information — they disagree about how it maps to policy.
A related reinterpretation: Jarociński and Karadi’s “information shock” component (JK_Info), which comoves with stocks , is itself strongly predicted by pre-announcement financial stress in Chen’s data — suggesting it reflects the Fed’s response to financial conditions rather than exclusive Fed knowledge.
Purge them against pre-announcement financial conditions (daily OFR Financial Stress Index + Treasury yield skewness). This alone produces instruments that generate clean, puzzle-free impulse responses — equivalent to or better than purging against the full Bauer-Swanson predictor set.
The identification problem is well-known. Gertler and Karadi use high-frequency surprises as external instruments in a Proxy SVAR , but Caldara and Herbst show that failing to account for the Fed’s systematic response to credit spreads attenuates estimated monetary policy effects . Bauer and Swanson’s solution is to orthogonalize MPS against six pre-announcement predictors (yield curve slope, S&P 500, commodity prices, employment growth, nonfarm payroll surprise, Treasury skewness) .
Chen (2026) shows that orthogonalizing Nakamura-Steinsson surprises against just two daily financial variables yields impulse responses free of price and output puzzles — and in fact more conventional at short horizons than the Bauer-Swanson-orthogonalized version . This is what the paper terms financial-conditions-sufficiency: once financial information is purged, additional economic predictors add little.
Practical recipe:
No — the Fed adopts a “wait-and-see” approach for data released within roughly two weeks of the meeting, fully incorporating only data released three or more weeks prior. Markets misread this as aggressive responsiveness.
The dominant view, formalized by Cieslak and Bauer and Swanson , is that markets systematically underestimate the Fed’s response to economic news, producing positive co-movement between pre-announcement economic surprises and policy surprises. Schmeling, Schrimpf and Steffensen similarly document expectation errors consistent with underreaction .
Chen (2026) finds the opposite sign once financial conditions are controlled: a positive real activity surprise in the two weeks before an FOMC meeting predicts a dovish policy surprise, not hawkish . This reverses the sign predicted by the “response to news” hypothesis and identifies what the paper calls the wait-and-see channel.
Timing evidence (Chen 2026):
Implication for identification: If you’re running event studies around unscheduled meetings, control for recent real activity surprises alongside financial conditions. The wait-and-see effect is concentrated there.
No — the empirical pattern runs the wrong way. Risk premia respond to monetary policy surprises after the announcement, rather than generating them.
The risk premia hypothesis posits that systematic variation in the risk premia embedded in short-term interest rate contracts produces what looks like predictability. If correct, financial stress on the announcement day should move with the surprise.
It doesn’t. Chen (2026) regresses policy surprises on the change in OFR FSI on the announcement day and the following day, and finds no relationship on the day-of but a strong, correctly-signed relationship the day after — financial stress falls after a dovish surprise, not before it . The FSI barely moves on FOMC days themselves.
This aligns with prior skepticism. Bauer and Swanson argue the required risk premia variation is implausibly large , and Piazzesi and Swanson show fed funds futures risk premia are small . It also fits the broader literature documenting policy-to-risk-premia transmission: Bernanke and Kuttner on equity reactions , Hanson and Stein on long rates , and Drechsler, Savov and Schnabl on the risk-taking channel .
Bottom line: Risk premia are a consequence of policy surprises, not their source.
Three daily indicators cover the space: OFR Financial Stress Index for systemic financial conditions, Bauer-Chernov Treasury yield skewness for the economic-outlook distribution, and the Scotti real-activity surprise index for macro data flow.
High-frequency FOMC event studies have long suffered a trade-off. Miranda-Agrippino and Ricco address information insufficiency with dynamic factor models on monthly macro data , but monthly data can’t be causally linked to irregular meeting dates. Chen (2026) argues a daily, information-rich combination resolves this .
The three measures:
Alternatives and caveats: The Gilchrist-Zakrajšek excess bond premium works as a robustness check for the FSI (Chen 2026 confirms results replicate). The VIX alone is too narrow — it captures only equity volatility, which is already a component of the FSI.
Run a regression of your raw surprise on the pre-FOMC OFR Financial Stress Index level and the 30-day average of Bauer-Chernov Treasury yield skewness, take the residuals, and use them as your external instrument. Chen (2026) shows this two-variable purge produces puzzle-free impulse responses equivalent to or better than the six-variable Bauer-Swanson purge .
Concrete recipe:
surprise ~ FSI_t-1 + TreasurySkew_t-30:t-1 via OLS; save residuals.Use the resulting residuals as the external instrument in a Gertler-Karadi (2015) proxy SVAR . Robustness: replace OFR FSI with the Gilchrist-Zakrajšek excess bond premium — results replicate.
Related questions: Where do I download FSI and Treasury skewness data? · Does this purge work for ECB and BoE surprises?
Three sources cover the full toolkit needed to replicate or extend Chen (2026) , all publicly available and freely downloadable.
For the raw surprises themselves: Bauer-Swanson MPS and Nakamura-Steinsson are available at the authors’ websites; Jarociński-Karadi from the AEJ:Macro data archive. For ECB equivalents, use the Altavilla et al. Euro Area Monetary Policy Event-Study Database (EA-MPD).
Related questions: How does the purge differ for unscheduled vs. scheduled meetings? · Does the OFR FSI work as a robustness check against EBP?
Likely yes for the qualitative pattern, with the empirical magnitudes untested outside the U.S. Chen (2026) is U.S.-only , but the structural argument — central banks respond to financial conditions to hit their economic targets, markets miss this channel — is not U.S.-specific. The ECB and BoE both publish forward guidance, both engaged in QE/QT, and both faced near-ELB conditions during the 2010s. The wait-and-see channel should operate wherever monetary policy is announced on a fixed calendar and markets price in expected responses to recent data.
The natural empirical extension uses the Altavilla et al. Euro Area Monetary Policy Event-Study Database for ECB surprises and the Cesa-Bianchi, Thwaites, Vicondoa (2020) UK monetary surprises for the BoE. Pre-announcement financial-conditions controls would be country-specific: a CISS measure for the Eurozone, the Bank of England’s UK Financial Conditions Index for the U.K.
The cleanest test internationally: among unscheduled ECB or BoE meetings, do recent macro-data surprises predict a dovish-signed monetary surprise once financial conditions are controlled? If yes, the U.S. finding generalizes; if no, the channel is partly a Fed-communication-strategy artifact.
Related questions: How do I download EA-MPD data? · Does the wait-and-see channel survive in unscheduled-meeting samples?
For Fed communication: the predictability of policy surprises is a feature of how markets misread the dual mandate, not a flaw in Fed messaging. Chen (2026) argues that markets take the “we don’t target financial conditions” statement literally and miss the channel — the gap is structural and not closed by market learning .
For market practitioners, three actionable implications:
Related questions: Does the Fed have private information about the economy? · What does the response-to-news hypothesis miss?
All data and code for Chen (2026) are available at robinchen.org . The paper uses:
Chen, Zhengyang. 2026. “Demystifying Monetary Policy Surprises: Fed Response to Financial Conditions and Wait and See for New Economic Data.” Journal of Macroeconomics 87: 103736. https://doi.org/10.1016/j.jmacro.2025.103736
@article{chen2026demystifying,
title={Demystifying Monetary Policy Surprises: Fed Response to Financial Conditions and Wait and See for New Economic Data},
author={Chen, Zhengyang},
journal={Journal of Macroeconomics},
volume={87},
pages={103736},
year={2026},
publisher={Elsevier},
doi={10.1016/j.jmacro.2025.103736}
}