Literature Readings · Prediction Markets · Paper C Proposal
Does the Federal Reserve Listen to Kalshi?
Textual and High-Frequency Evidence on Central Bank Use of Prediction-Market Information
Working draft research proposal · May 19, 2026 · Does the Fed listen, react, or merely watch?
📋 Executive Summary
Diercks-Katz-Wright (NBER 34702, Jan 2026) established that Kalshi macro markets beat Bloomberg consensus and Fed Funds Futures on FOMC rate-decision forecasting. Paper C asks the natural next question: does the Fed know this, and does it change Fed behavior?
Three sub-channels: (i) Reading — FOMC minutes / transcripts / staff memos / speeches for prediction-market language (text mining 1994-2026); (ii) Reaction function — does Kalshi probability of a higher-rate decision enter the Fed's information set as a state variable beyond what Fed Funds Futures already span?; (iii) Convergence — when Kalshi and Fed Funds Futures diverge before FOMC, does the actual decision converge to Kalshi or to FFF?
Coauthor strategy: direct outreach to Diercks, Katz, or Wright (the NBER 34702 authors) for institutional access + legitimacy. Target: AER / J Monetary Economics / J Finance. Timeline: 18 months for the bundle (Paper C is the final installment); ~60% of pipeline reused from Papers A and B. Side product: the FOMC text-mining + minute-level Kalshi-FFF identification toolkit becomes the canonical "prediction-market information in Fed policy" methodology.
1. Research Question
Q: Does the Federal Reserve incorporate Kalshi prediction-market pricing into its information set or reaction function?
Three sub-questions, in order of increasing causal aggressiveness:
Q1 (Reading — descriptive). Do FOMC officials and Fed staff mention Kalshi, "prediction markets," or "event contracts" in minutes, transcripts (5-year-lag), speeches, working papers, and presentations? When? In what context?
Q2 (Reaction function — empirical). Does Kalshi's expected-rate-path enter the Fed's reaction function as an information variable beyond what is already spanned by Fed Funds Futures, the Survey of Professional Forecasters, Bloomberg consensus, and standard macro indicators?
Q3 (Convergence — causal). In pre-FOMC windows when Kalshi and Fed Funds Futures diverge, does the Fed's actual decision converge to Kalshi pricing or FFF pricing?
The single headline statistic we are after:
A "Kalshi loading" on the Fed reaction function — how much weight the Fed places on Kalshi probabilities relative to the standard information set. If 0 → Fed ignores Kalshi. If > 0 → Fed reads Kalshi. If >> 0 and convergence is to Kalshi → Fed materially relies on Kalshi.
2. Positioning vs Existing Literature
| Paper / Strand | What they did | What we add |
|---|---|---|
| Diercks-Katz-Wright NBER 34702 (Jan 2026) | Documents Kalshi's perfect FOMC forecast record. Argues Kalshi could be a useful Fed information source. | Tests whether Kalshi is a Fed information source. Their paper is the necessary first step; ours is the necessary second. |
| Bauer & Swanson AER 2023 "An Alternative Explanation for the 'Fed Information Effect'" | High-frequency identification of monetary policy shocks. Decomposes "Fed information effect" into surprise vs response. | Methodological template — same identification logic for Kalshi-FOMC convergence. |
| Nakamura & Steinsson QJE 2018 | Identifies "Fed information effect" from FOMC announcement surprises in fed funds futures. | Same logic but with Kalshi as the alternative information variable. |
| Hansen-McMahon JIE 2016 + others on LDA-based FOMC text analysis | LDA topic models of FOMC minutes; finds different topics correlate with different policy outcomes. | Extension to prediction-market language. Use Hansen-McMahon LDA pipeline as a starting point. |
| Romer-Romer narrative monetary policy AER 2004 | Manual reading of FOMC minutes to identify monetary policy intentions. | Read for prediction-market mentions explicitly + use LDA for automated detection. |
| Greenwood-Hanson-Stein JF 2010 | "A Comparative-Advantage Approach to Government Debt Maturity" — Treasury responds to market signals. | Same logic but for Fed responding to Kalshi rather than Treasury responding to bond markets. |
| Smith-Timmermann-Wright 2024+ working papers on market-based macro forecasting | Market prices vs survey-based forecasts comparison in macro. | We test which forecast the Fed actually USES vs which is most accurate. |
One-paragraph positioning: Diercks-Katz-Wright (Jan 2026) established Kalshi's forecasting accuracy. Bauer-Swanson (2023) + Nakamura-Steinsson (2018) provide the methodological template for testing whether the Fed uses market-based information. Hansen-McMahon (2016) provides the LDA toolkit for FOMC text analysis. Our paper combines these into a direct test of whether the Fed reads, reacts to, and converges with Kalshi. No paper yet bridges Kalshi to Fed reaction-function literature.
3. Data Architecture
3.1 Kalshi macro data
Same Kalshi data pipeline as Paper A but extended back to 2022 launch + forward through 2026. Focus on:
- KXFED contracts (FOMC rate decisions) — ~30+ events 2022-2026
- KXCPI (monthly CPI prints) — ~50 events
- KXNFP (non-farm payrolls) — ~50 events
- KXRECSSNBER (recession-call markets)
- KXGDP (quarterly GDP prints)
3.2 FOMC textual data (free, Fed public)
| Minutes | Released ~3 weeks post-meeting. 1993-present. Public download from federalreserve.gov |
| Transcripts | Released with 5-year lag. Full verbatim discussion. Allows pre-Kalshi vs post-Kalshi within-FOMC analysis |
| Speeches | All Fed governor speeches 2000-present, ~500/year |
| Staff memos | Tealbook / Greenbook materials (5-year-lag) referenced in minutes |
| SEP dot plots | Quarterly Summary of Economic Projections |
3.3 Fed Funds Futures (CME, intraday via WRDS)
CME 30-day Fed Funds Futures with 1-minute intraday data. Identification: changes in implied rate from pre-FOMC to post-FOMC announcement = monetary policy shock (à la Kuttner 2001, Gürkaynak-Sack-Swanson 2007). Cross-reference with Kalshi at same minute.
3.4 Survey of Professional Forecasters
Philadelphia Fed SPF (quarterly) — the "consensus" benchmark. Compare PolyMarket forecasts to SPF on rate decisions, inflation, GDP, unemployment.
3.5 Treasury auction data
Treasury Direct: auction date, security, bid-cover ratio, tail (difference between awarded yield and pre-auction WI yield), foreign demand. Test: pre-auction Kalshi rate pricing should predict auction outcomes if institutional buyers use Kalshi information.
3.6 SEP dot plot revisions
Quarterly SEP at March, June, September, December. Compare dot-plot median for end-of-year rate to Kalshi end-of-year rate distribution at the same date. Test: Fed dot plot revisions in Q to Q+1 should be correlated with Kalshi changes if Fed updates beliefs based on Kalshi.
4. Identification Strategy
Two main identification approaches, complementary:
4.1 Pre-FOMC convergence test
In the 30 minutes before an FOMC announcement, Kalshi and Fed Funds Futures may diverge in implied probability of a rate move. The realized Fed decision then converges to one or the other — or neither.
Hypothesis: β > 0 → Fed leans toward Kalshi when they disagree with FFF. β = 0 → Fed indifferent. β < 0 → Fed leans toward FFF.
Identifying assumption: When Kalshi and FFF disagree, the disagreement is informationally meaningful (Kalshi has private information not in FFF) rather than purely noise. Test by examining whether pre-FOMC Kalshi-FFF spread predicts post-announcement market reactions.
4.2 Reaction function estimation (Taylor-rule-augmented)
Standard Taylor rule + Kalshi as additional regressor:
where it is the fed funds rate, Kalshit is the pre-FOMC Kalshi rate expectation. Test: H0: θK = 0 (Kalshi doesn't matter beyond standard variables). HA: θK ≠ 0.
Identification: Kalshi enters the reaction function only via its information content beyond π, ỹ, and lagged i. We control for the SPF and FFF; the remaining Kalshi effect is the marginal information.
4.3 FOMC text analysis
Build a "prediction-market mention" indicator at the FOMC-meeting level using:
- Explicit keyword search: "Kalshi", "prediction market", "event contract", "PolyMarket"
- LDA topic modeling: identify a "market-based forecasting" topic and measure its prominence over time
- Named-entity extraction: which staff members raise PM-related issues?
- Contextual classification: when PMs are mentioned, is it as a source of information, a source of noise, or a regulatory concern?
Hypothesis: Mentionsi should be increasing over time (post-2022). Should correlate with Kalshi-FFF divergence (Fed discusses PMs more when they disagree).
5. Main Specifications
5.1 Spec 1 — Convergence test (Q3)
Run §4.1 convergence regression across all FOMC meetings since Kalshi launched (2022 onward, ~30 meetings). Test sign and significance of β.
5.2 Spec 2 — Augmented Taylor rule (Q2)
Run §4.2 augmented Taylor rule. Compare θK estimate to:
- θSPF (does Kalshi enter beyond SPF?)
- θFFF (does Kalshi enter beyond FFF?)
- Horse-race: regress on Kalshi, SPF, FFF jointly and test which dominates
5.3 Spec 3 — Granger causality (Q3)
Minute-resolution within pre-FOMC windows:
Test which direction is statistically stronger. If FFF Granger-causes Kalshi but not vice versa → Kalshi is purely lagging FFF. If both → bidirectional causality. If Kalshi Granger-causes FFF → Kalshi has private information FFF doesn't have.
5.4 Spec 4 — FOMC text time-series (Q1)
Plot Mentionsi over time. Document:
- When did first explicit Kalshi mention appear?
- What was the context (mentioned positively, neutrally, or skeptically)?
- Cross-section by FOMC member: who mentions PMs most?
- Correlation between Mentionsi and Kalshi-FFF divergence in FOMCi
5.5 Spec 5 — Treasury auction outcomes
For each Treasury auction (weekly bills, monthly notes), test:
where Taili is the difference between awarded yield and pre-auction WI yield. Hypothesis: If institutional buyers use Kalshi information, pre-auction Kalshi pricing should predict auction outcomes (positive β).
5.6 Spec 6 — Dot plot revisions
Quarterly SEP dot plots. Test:
where ΔDotq is the median end-of-year rate forecast change in SEP from quarter q-1 to q. Hypothesis: Fed should update dot plots in direction of recent Kalshi moves if Fed integrates Kalshi.
6. Coauthor Strategy
This paper benefits especially from a Fed-economist coauthor. Three approaches:
Option A — Approach Diercks-Katz-Wright directly. They wrote the Jan 2026 NBER paper. Pitch: "we want to extend your work — would you coauthor or comment?" Best legitimacy outcome. Risk: they may want to do this themselves.
Option B — Approach a different Fed PhD economist. Many Fed economists are sympathetic to this kind of work. Approach via NBER conference circuit (Asset Pricing, Monetary Economics).
Option C — Solo with strong references. No Fed coauthor; rely on Diercks-Katz-Wright as cited inspiration. Doable but lower visibility.
Recommendation: Try Option A first. The Diercks-Katz-Wright team is the natural fit, and their inclusion would dramatically improve the paper's credibility and access. If they decline, fall back to Option C with active outreach to NBER Monetary Economics Program.
7. Robustness
- Alternative Kalshi measures: probability of next-FOMC rate cut vs probability of rate cut in 3 months vs probability of recession in 12 months
- Subsample tests: 2022-2023 (Kalshi launch period, low salience) vs 2024-2026 (high salience)
- Cross-FOMC heterogeneity: some FOMC meetings have larger Kalshi-FFF divergences; do those drive results?
- Alternative reaction-function specs: forward-looking vs backward-looking Taylor rule; nominal-GDP-targeting variant
- Different LDA topic numbers: 10, 25, 50, 100 topics — robust to choice?
- Manual coding of FOMC minutes: 30 minutes from 2022-2026 manually classified to validate LDA
- SEP dot plot heterogeneity: do certain SEP members revise more in direction of Kalshi?
- Excluding crisis periods: drop COVID 2022-23 era; recheck
8. Falsification
F1 — Pre-Kalshi placebo. Run convergence test on FOMC meetings 2010-2021 using PredictIt rate-decision markets instead of Kalshi. PredictIt was thinly traded for macro markets — should show much weaker convergence. If convergence appears anyway, our identification is spurious.
F2 — Irrelevant Kalshi markets. Use Kalshi sports markets or weather markets as the Kalshi variable in the reaction function. They should have θ ≈ 0. If they don't, we're picking up generic time-varying noise.
F3 — Cross-country. Test whether the ECB / BoE / BoJ similarly correlate with Kalshi macro contracts. ECB / BoE etc. should NOT correlate strongly with US Kalshi (they don't read it). If they do, common-trend confound is present.
F4 — Pre-launch placebo. "Mention" indicator should be ≈ 0 for FOMC minutes before 2022 (Kalshi launch). If we find pre-2022 mentions, our keyword matching is too broad.
9. Risks & Mitigations
| Risk | Severity | Mitigation |
|---|---|---|
| Null finding: Fed doesn't read Kalshi | Medium | Null is still publishable! "Fed ignores Kalshi despite Diercks-Katz-Wright accuracy" is a striking finding. Frame appropriately. |
| Small sample: only ~30 FOMC meetings post-Kalshi-launch | High | Augment with CPI/NFP/GDP-print events (~150 total). Use Bayesian shrinkage / cross-event pooling. |
| FFF and Kalshi may be too similar (highly correlated) | Medium | Focus on divergence periods (Kalshi-FFF gap > 5pp). ~10-15 such periods per year provide identification. |
| Diercks-Katz-Wright may publish follow-up themselves | Medium | Approach them early to coauthor (Option A above). If they want to do it alone, we shift focus to the broader cross-asset implications. |
| LDA topic modeling may not find a clean "PM topic" | Medium | Use multiple methods: keyword search (always works) + LDA + manual coding of validation sample. |
| 5-year transcript lag means most 2024-2026 transcripts unavailable | Medium | Use minutes (3-week lag) + speeches as proxy for 2024-2026. Detailed transcript analysis on 2022-2020 baseline. |
| Fed-coauthor outreach fails | Medium | Solo version is feasible. Methodology is standard enough that Fed-economist coauthor is desirable but not essential. |
| Reverse causality: Fed announcements move Kalshi, not vice versa | High | Focus on PRE-FOMC windows. Granger causality test directly addresses this. |
10. Timeline & Resources
| Phase | Month | Deliverable |
|---|---|---|
| Coauthor outreach | 1 | Approach Diercks-Katz-Wright; backup outreach to NBER Monetary Economics community. |
| Data collection | 1-2 | FOMC minutes/transcripts/speeches scraping (free, automated). Kalshi macro data (already in pipeline from Paper A). |
| FFF intraday | 2 | WRDS download of CME Fed Funds Futures minute-resolution 2022-2026. |
| Text analysis | 3-4 | Keyword indicator + LDA + manual coding validation. |
| Convergence + reaction function | 4-5 | Specs 1, 2, 3 (convergence, augmented Taylor, Granger). |
| Treasury auction + dot plot | 5-6 | Specs 5, 6. |
| First draft | 6-7 | Working paper ready. |
| Polish + submit | 7-8 | Submit to AER (or QJE / J Monetary Economics). |
Note: Paper C timeline starts ~12 months after Paper A's data pipeline is built. Total bundle: Paper A Month 6-9, Paper B Month 12-15, Paper C Month 15-18.
11. Policy & Institutional Impact
Paper C has the most direct institutional impact — it informs how a top-tier central bank uses (or doesn't use) market-based information:
- Fed Board of Governors: direct policy implication. If we show Kalshi is uniquely informative and the Fed reads it, the Fed gains institutional cover for citing PMs. If we show Fed doesn't read Kalshi but should, that's a recommendation.
- Treasury Department: if Kalshi predicts auction outcomes, Treasury could use Kalshi for optimal debt-maturity decisions (Greenwood-Hanson-Stein logic applied).
- CFTC: regulatory question — should Kalshi macro contracts be treated as "informational infrastructure" with different regulatory treatment than political event contracts?
- Other central banks: ECB, BoE, BoJ may follow our methodology to assess their own use of PM-equivalent information.
- Academic / NBER conference circuit: Monetary economics / asset pricing programs would find this highly placed.
The institutional politics: The Fed is sensitive about citing private-market information. If our paper shows the Fed already reads Kalshi (even informally), it provides cover for more explicit institutional adoption. If it shows the Fed ignores Kalshi despite accuracy, it's a constructive critique. Either way, the Fed will read this paper.
Working draft research proposal · Generated May 19, 2026