Measuring FOMC Communication Shocksa comparison of the dictionary and transformer-based text methods

Resumo

This paper compares alternative measures of FOMC communication based on a dictionary index, a fine-tuned BERT classifier, and a fine-tuned FinBERT classifier. Each measure is orthogonalized with respect to a conventional rate surprise and related to intraday movements in U.S. Treasury yields and equity prices around FOMC announcements using local projections and pairwise encompassing regressions. The results show that the relative performance of the methods depends on the asset class. For the 2-year Treasury yield, the dictionary-based shock contains information not captured by generic BERT. For the S&P 500, by contrast, the BERT-based shock encompasses both the dictionary and FinBERT measures across the horizons considered. Overall, the findings suggest that the choice of textual methodology matters for empirical conclusions about FOMC communication, and that no single approach uniformly dominates across asset classes.

Ficha do documento

Tipo

Dissertação

Ano

2026

Instituição

Fundação Getulio Vargas

Fonte

Repositório da FGV

Idioma

Inglês

Acesso

Acesso aberto

Identificador

oai:repositorio.fgv.br:10438/40854

Temas

EconomiaDesenvolvimento Econômico

Palavras-chave

Comunicação de banco centralChoque monetário textualIdentificação de alta frequênciaCentral bank communicationText-based monetary shocksHigh-frequency identificationEconomiaPolítica monetária - Estados UnidosMercado financeiroDesenvolvimento econômicoComitê Federal de Mercado Aberto