The four-day concept map
Only if messages alter expectations and remain credible relative to mandate and instruments.
No—not directly. Surveys, prices, and text are traces with structured wedges.
Use narrow windows and market instruments to separate expectations from surprises.
Separate publication, exposure, interpretation, belief revision, and behaviour.
Questions by lecture
Lecture 1 · Communication as policy
- What is the mechanism from a central-bank message to an economic decision?
- Why is transparency neither always beneficial nor identical to credibility?
- How do risk and Knightian uncertainty change the communication problem?
- What does it mean for inflation expectations to be anchored?
- Why are independence and accountability complements as well as tensions?
- When does forward guidance describe a reaction function, and when does it resemble a commitment?
Lecture 2 · Measuring expectations
- Specify an expectation completely: agent, variable, horizon, moment, jurisdiction, and probability measure.
- What does a survey observe that a market price does not—and vice versa?
- Why is a futures-implied rate not automatically a physical expectation?
- Why does a meeting-month futures contract require a calendar-weight adjustment?
- Which wedges could explain a disagreement between survey and market measures?
- How can triangulation improve measurement without pretending one source is a gold standard?
Lecture 3 · Monetary-policy surprises
- Why does the realised policy-rate change not measure the policy surprise?
- What economic news is captured by the target factor and by the path factor?
- Why is the choice of event window an identification assumption?
- How can a central-bank information shock contaminate a conventional policy shock?
- What changes when the foreign market is closed during an FOMC announcement?
- How does the interpretation narrow as you move up the estimand ladder?
Lecture 4 · Communication as text data
- Why is a published document not the same thing as an observed communication treatment?
- How do bundled releases complicate attribution?
- What is the difference between unexpected tone and the level of tone?
- How would you measure public exposure rather than institutional publication?
- Why can staged disclosure change both attention and interpretation?
- How should domain knowledge enter dictionary, embedding, or supervised text measures?
Working glossary
Expectation
A belief about a future variable conditional on an information set. It is incomplete unless the agent, horizon, moment, and probability measure are specified.
Anchoring
The degree to which longer-run expectations remain stable when near-term news arrives. Stability, level, and sensitivity are different empirical dimensions.
Risk-neutral vs physical
Market prices often reveal expectations under a pricing measure. The physical expectation relevant for forecasting can differ because risk is priced.
Measurement wedge
A systematic force separating an observed proxy from the target concept—for example risk premia, calendar averaging, liquidity, or survey selection.
Event window
The interval over which asset-price changes are attributed to an announcement. Narrower windows reduce confounding but may miss slow adjustment.
Target factor
The component of announcement news concentrated in the current policy-rate decision or very near-term expected rate.
Path factor
News about the expected trajectory of policy beyond the current decision, extracted from changes across multiple interest-rate instruments.
Information shock
News markets infer from the central bank’s private assessment of the economy, potentially moving policy expectations and growth-sensitive prices together.
Exposure
Whether an audience actually encounters a communication. Publication is supply; exposure also depends on media, timing, attention, and access.
Unexpected tone
The component of textual stance not predicted by the economic state, document type, speaker, or other pre-release information.
Common mistakes to avoid
Rate change = shock
A scheduled and fully expected rate change contains no surprise. Identification requires the unexpected component.
Price = belief
A market price combines beliefs with risk pricing, liquidity, institutional constraints, and contract design.
Published = received
A document can be public without being noticed, understood, believed, or acted upon by the audience of interest.
Narrow = causal
A narrow window improves attribution, but simultaneity, information effects, anticipation, and market microstructure can remain.
One factor = one message
Policy announcements bundle current decisions, future guidance, forecasts, risks, and institutional signals.
Sentiment = meaning
Generic positive or negative language does not automatically map into hawkishness, stance, risk, clarity, or policy commitment.
Final revision checklist
- I can draw the communication-to-outcomes chain and locate attention, beliefs, and behaviour.
- I can explain why credibility depends on institutional backing rather than rhetorical confidence alone.
- I can define the expectation object before choosing a proxy.
- I can list the principal wedges in survey and market-based measures.
- I can explain the meeting-month futures calendar-weight problem intuitively.
- I can distinguish a policy action from a target surprise and a path surprise.
- I can defend an event window and identify likely contaminants.
- I can explain the central-bank information effect.
- I can trace a domestic surprise into a global spillover design.
- I can separate text, release, exposure, interpretation, and outcome.
- I can propose at least one validation exercise for a text measure.
- I can state the estimand and the assumptions required for its causal interpretation.