Hawkometer — Central Bank Speech & Communication Sentiment Tracker
Score every Powell, Lagarde, Bailey and Ueda appearance on a hawkish/dovish scale, in near real time
Updated September 1, 2026Why a Hawkometer?
Markets do not just react to rate decisions — they react to the language around them. A single phrase from the Fed Chair, an interview with the ECB’s chief economist, or a testimony from the Governor of the Bank of England can move yields by basis points before the next meeting is even on the calendar.
The Hawkometer is a structured, machine-readable index of that language. Every public appearance by a senior official at one of nine major central banks is scored on a single scale:
- +10 — most hawkish (clear signal that policy will tighten or stay restrictive)
- 0 — neutral
- -10 — most dovish (clear signal that the easing cycle is approaching or under way)
Scores are aggregated upward into speaker, committee and cross-bank views, and combined with a 30-day vs. prior-60-day shift indicator that flags when a committee’s tone is turning.
Powered by AI analysis
As of June 2026, the Hawkometer uses Claude AI (by Anthropic) to score each speech — moving beyond static keyword matching to context-aware reasoning that understands negation, hedging, and the subtle shift signals that drive markets. Every score includes a written rationale so the analysis remains auditable. How it works →
The global dashboard
The bar chart below ranks each committee by its current 90-day, voter-weighted Hawkometer score.
A score above zero means that, over the last 90 days, the weighted average of all speeches by that committee has tilted hawkish. A score below zero means it has tilted dovish.
Per-bank Hawkometers
Each central bank has a dedicated Hawkometer page covering committee composition, voting bloc lean, recent speech sentiment and the upcoming calendar of appearances.
Coverage
| Bank | Committee | Speakers tracked |
|---|---|---|
| Federal Reserve | FOMC | Chair Powell, Vice Chairs, Governors and the regional Fed Presidents on the rotating schedule |
| European Central Bank | Governing Council | President Lagarde, Executive Board, Bundesbank, Banque de France, DNB, Banca d’Italia |
| Bank of England | Monetary Policy Committee | Governor Bailey, Deputy Governors, Chief Economist, External members |
| Bank of Japan | Policy Board | Governor Ueda, Deputy Governors, Policy Board members |
| Reserve Bank of Australia | Board | Governor Bullock, Deputy Governor, Treasury Secretary, Assistant Governor |
| Bank of Canada | Governing Council | Governor Macklem, Senior Deputy Governor, Deputy Governors |
| Reserve Bank of India | MPC | Governor, Deputy Governor, External members |
| Swiss National Bank | Governing Board | Chairman Schlegel, Vice Chairman, Member |
| People’s Bank of China | Governor and key deputies | Governor Pan Gongsheng, Deputy Governors |
How to read the index
A score is a summary, not a verdict. Two observations are worth keeping in mind:
- Direction matters more than absolute level. A committee at +1.5 that was at +3.0 a month ago is moving dovish, even if the headline reading still looks hawkish. The shift indicator captures this.
- The Chair sets the centre of gravity. A hawkish regional Fed President paired with a dovish Chair often nets out to neutral. We weight by voting status, not by office, so the Chair’s vote counts the same as any other voter — but their public language tends to anchor expectations more than other speakers.
For a full description of the scoring engine, the phrase library and known limitations, see the methodology page.
Hawks, Doves, and Why a Single Speech Can Move Markets
Central bankers are sorted into two informal camps by the way they talk about monetary policy. Hawks worry first about inflation. They favor higher rates, tighter policy, and are quick to flag wage pressures or rising prices as reasons to act. Doves worry first about unemployment and growth. They prefer lower rates, are more comfortable letting inflation run a little above target if it means more jobs, and emphasize the costs of cooling the economy too aggressively. Most officials slide along a spectrum rather than sitting at either extreme, and the same person can shift hawkish or dovish depending on the data.
When a Federal Reserve governor gives a speech, traders read it as a clue to where the committee is heading. If several members sound more hawkish in the same month, markets may raise the probability of a rate hike, bond prices can fall, and the dollar can strengthen.
Not every official's vote counts equally at every meeting. Voting members usually move markets more sharply, but non-voters still matter because they join the debate and may rotate into voting seats later. The Hawkometer weights speakers accordingly.
Committee Dispersion, Voting Mechanics, and Sentiment Scoring Methodology
Cross-committee dispersion in revealed preferences is itself a tradable signal. The FOMC, ECB Governing Council and BoE MPC reveal disagreement through different institutional channels: dot plots, national central bank speeches, consensus press conferences and published vote splits. Non-voting member speeches can contain substantial signal during transition regimes because rotation reveals future voting preferences.
Sentiment scoring approaches range from dictionary methods and custom hawk-dove phrase libraries through topic models to fine-tuned transformer models. Keyword scorers like the Hawkometer offer full traceability because every score decomposes into the matched phrases that produced it, while context-aware models can capture nuance at the cost of lower auditability.
Committee scores at a glance
Voter-weighted 90-day rolling Hawkometer score for each committee. Hover any card for the underlying speaker breakdown.
Which committee is most hawkish right now?
Latest scored appearances
About this index
The Hawkometer uses Claude AI (Anthropic) to score central bank communication with context-aware reasoning. Every score is accompanied by a written rationale and the key phrases that drove it. For full details on the scoring engine, voter weighting, rolling windows and known limitations, see the methodology page.