Description
Central-bank communication does more than signal policy stance: it also explains why inflation rises, falls, or persists. Yet computational work on central-bank text has concentrated mainly on tone, topics, uncertainty, or market-relevant news, leaving causal attribution comparatively underexamined. This study traces explicit inflation-causal constructions in a longitudinal corpus of Federal Reserve, European Central Bank (ECB), and Bank of England (BoE) speeches. The confirmatory sample contains 3,950 monologic speeches from a common Bank for International Settlements sampling frame spanning 1999 to July 2026. A transparent construction-first detector identified a broad layer of 3,807 inflation-plus-causal-cue sentences and a stricter directional layer of 2,177 sentences in which causal form and inflation expression occurred in a constrained cause–effect configuration. Negative-binomial models with token exposure, genre and period controls, and speaker-clustered standard errors show that the strongest result is temporal rather than institutional. The findings suggest that central-bank causal explanation is strongly state-responsive: when inflation becomes unusually salient or contested, policymakers devote substantially more linguistic space to causal accounts. They also show that apparent cross-institution differences are sensitive to extraction stringency.
Thông tin các tác giả
Nguyễn Thị Hồng Liên: ThS, đang công tác tại Trường Đại học Ngoại ngữ - Tin học TP. HCM, 828 Sư Vạn Hạnh, Quận 10, TP. HCM, email: liennth1@huflit.edu.vn
Từ khóa
central bank communication; inflation; causality extraction; causal attribution; computational linguistics; corpus linguistics
