Dragos Gorduza, Livia Silva Paranhos and Ivan Yotzov argue in Bank Underground about MPC transcripts

Dragos Gorduza, Livia Silva Paranhos and Ivan Yotzov argue in Bank Underground about MPC transcripts

Published on October 15, 2026, Dragos Gorduza, Livia Silva Paranhos and Ivan Yotzov use large language models (LLMs) to analyse Bank of England Monetary Policy Committee (MPC) transcripts, extracting measures of members’ views on inflation and growth and the arguments behind policy decisions. They argue that vote tallies can conceal different concerns behind the same decision. In their analysis, growth concerns weakened and later strengthened over the sample, while concerns about above-target inflation increased. They report that both measures predict votes for higher Bank Rate, but only growth measures explain votes to cut it. They also report that these measures relate to voting beyond aggregate data and official forecasts. Their analysis focuses on Bank Rate votes, not asset purchase decisions. The authors conclude that historical transcripts contain information beyond vote tallies and suggest comparing transcript narratives with contemporaneous public communications as future work. Whether their measures capture policymakers’ full reasoning remains unknown.

Publication: d7104557-618a-d0ad-d265-98eea126e7c2

attribution — pass: The reading correctly attributes the analysis to Dragos Gorduza, Livia Silva Paranhos and Ivan Yotzov on Bank Underground. It keeps findings as the authors' reported results rather than Bank of England policy, which matches the blog's disclaimer.

The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

summary — uncertain: The summary never says the transcripts cover 2015 to 2017, a period that includes the June 2016 EU referendum. It says concerns changed 'over the sample' without defining that sample, so a reader cannot tell when these shifts happened or that the transcripts are historical and released with an eight-year lag. 'Bank Rate' is also central but not explained as the Bank of England's main policy interest rate. Smallest repair: name the 2015–17 period, the Brexit-referendum context, and gloss Bank Rate.

Our main data set comprises 258 publicly available statements from individual MPC members covering 29 meeting transcripts between 2015 and 2017.

date — pass: The reading states October 15, 2026, which matches the UTC day of originalPublished and the page dateline. The second date in the dateline (14/10/2026) is the page's own artifact and is not repeated. The page was edited after publication, but the cited passages are core article text. There is no hindsight framing, and the future work is framed as a suggestion.

15/10/202614/10/2026

support — pass: Each claim matches body text that is plainly part of the original post, so the later edit does not cast doubt on them. The 'beyond aggregate data and official forecasts' statement matches the regression controls the authors describe. The limitation about full reasoning is appropriately cautious, and the question follows from the authors' stated future work.

Therefore, we test whether these new metrics capture any variation in voting behaviour over and above the aggregate data and official forecasts.

media — pass: No media was selected and no verified candidates were supplied, so there is nothing to misidentify.