Research

Before an FOMC meeting, Federal Reserve staff prepare several complete statements, but only one becomes public and receives a market price. The rejected drafts record what the Committee could have communicated, not how markets would have understood it. I make these drafts comparable by using a large language model (LLM) to simulate 30 synthetic traders with varying risk tolerances and perspectives, each generating interest-rate forecasts for each statement. The disagreement among those forecasts measures how tightly the statement pinpoints the future policy path. Across 183 statements, greater disagreement is associated with larger immediate changes in short- and longer-term interest rates. In date-ordered tests, it also predicts which announcements will move markets most beyond premeeting conditions and the model’s average forecast. I then apply the procedure on 429 rejected drafts from 135 meetings. Across 79 meetings in which drafts proposed the same policy action, the FOMC committee favored those generating less disagreement about medium- and long-term rates. Drafts departing from the previous statement produce wider distributions, and the institutional evidence explains why. Staff pairs their expected action with inherited language; transcripts explain departures when policymakers change the future path or preserve flexibility. Together, the validated measure reveals economically meaningful differences among rejected statements, while the hidden menu shows that language is shaped as policy alternatives are constructed.

Literature on females in leadership positions in politics find positive effects on the electorate - health and education of children, public goods aimed at women, and women empowerment. In this paper we investigate how women political heads affect the mental well-being of the citizens. Using individual level health survey data from two large states of India, where the headship position for a fixed proportion of village councils is reserved for women and this allocation is randomised, we find that women electorates in reserved villages are happier relative to men in the unreserved villages. This result also bears out when we use self-reported quality of life as an indicator. We find that while men of all age groups are unhappy under a female leader, women aged 18-49 are happier than their unreserved counterparts, suggesting that women leaders may favour policies which affect gender relations. Mediation analyses point out that one of the mechanisms at play is that the women in reserved villages feel safer when on their own, a channel different from those found in the literature. Heterogeneity analyses point out that this impact on happiness is stronger for women in villages having a lower gender ratio. We also find that women higher up in the socio-economic ladder have the highest marginal effect.

Vagueness can reveal low confidence, depending on equilibrium sorting. A central bank privately knows the reliability of its assessment and commits, before observing its direction, to an unobserved disclosure probability. I characterize all pure-strategy perfect Bayesian equilibria. In the monotone separating equilibrium, a directional announcement raises perceived reliability and a non-directional announcement lowers it. Persistence can sustain this sorting although the low-reliability bank prefers greater precision in a one-shot game. A closed-form example produces separation for discount factors at least 1/3, alongside precise pooling. All messages occur under both protocols, so Bayes’ rule determines every belief.