TL;DR
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
A Swiss National Bank working paper by Severin Joho and Enzo Rossi analyzes 1,063 speeches by senior SNB officials from 1928 to 2024. It reports that speeches became more frequent, shorter and more accessible, while inflation gained prominence in policy discussions during the flexible exchange-rate era.
The paper, titled “The evolution of frequency, language complexity and policy narrative: benchmarking Python-based NLP and ChatGPT on a century of SNB rhetoric,” draws on a dataset of speeches delivered between 1928 and 2024. The authors examine how often officials spoke, the length and readability of their remarks, and the topics emphasized in them.
According to the study, annual speech frequency followed an upward trajectory from the mid-1950s, with a more pronounced acceleration after Switzerland moved to flexible exchange rates. The authors also report a clear shift toward shorter speeches and increasing accessibility in the language used since the 1960s.
The paper identifies a change in policy themes alongside those changes in delivery. Inflation became more prominent in the flexible exchange-rate era, while cyclical economic development and unemployment received less emphasis. The researchers also compare the outputs of probabilistic language models, including ChatGPT, with deterministic Python-based methods, and discuss methodological limitations and open questions.
How SNB Communication Changed
The findings offer a long-run view of how the Swiss central bank has communicated with public audiences as its policy environment changed. More frequent speeches and more accessible language may affect how officials explain policy and economic conditions; the paper documents patterns in the speeches but does not establish why each change occurred or how audiences responded.
The shift in thematic emphasis also matters for readers following central bank communication. The study reports that inflation took a more prominent place in speeches in the flexible exchange-rate era, while unemployment and cyclical developments received less attention. These are findings about the content of speeches, not direct evidence that the SNB’s policy decisions were determined by the language patterns.
The methodological comparison is relevant beyond this historical dataset. Researchers increasingly use generative AI to analyze large collections of texts. By testing ChatGPT against established Python-based text-mining approaches, Joho and Rossi address what such tools can add—and why their outputs require careful interpretation.
central banking and monetary policy textbook
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
A Century of SNB Speeches
The working paper examines official speeches across major changes in Switzerland’s monetary-policy setting. Its observation period begins in 1928 and ends in 2024, spanning decades before and after the transition to flexible exchange rates referenced by the authors.
Rather than relying on a single measure, the researchers assess speech frequency, length, readability and thematic content. They describe their dataset as a comprehensive collection of 1,063 speeches by high-ranking SNB officials. The paper’s historical approach makes it possible to compare communication patterns across different periods, though the source summary does not specify every category or coding decision used in the analysis.
The authors frame the comparison between language-analysis tools as a methodological part of the study, not just a technical detail. Python-based methods follow defined procedures, while generative models produce probabilistic responses to prompts. The paper says the approaches can complement each other, while also warning of algorithmic pitfalls.
“The frequency of annual speeches has followed an upward trajectory since the mid-1950s, accelerating notably after the transition to flexible exchange rates.”
— Severin Joho and Enzo Rossi, authors of the SNB working paper
natural language processing with Python book
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Limits of the Speech Analysis
The reported findings describe patterns in the speech collection; the source material does not provide detailed figures for the size of each change, such as year-by-year counts, average speech length or readability scores. It is therefore not possible from the summary alone to quantify the scale of the shifts or compare them across individual decades.
The summary also does not explain how the researchers attribute changes in speech frequency and topics to particular institutional or economic causes. The timing of the acceleration relative to the shift to flexible exchange rates is reported, but that association does not by itself establish causation. The authors flag algorithmic pitfalls and open research questions in applying large language models, and the available description does not provide their full evaluation results or detailed safeguards.
It remains unclear how closely ChatGPT’s findings matched the Python-based analysis on each task, and whether the authors found particular topics or periods more difficult to classify. Those details require the full paper rather than the summary supplied here.
As an affiliate, we earn on qualifying purchases.
Further Tests of Language Models
The working paper is available through the SNB research publications page. Its next steps are research questions rather than a stated policy action: the authors call for further work on how large language models can be used to study central bank communication, including the risks and limitations of those tools.
Readers seeking the methods and detailed results can consult the full paper, including its analysis of the dataset and comparison between ChatGPT and Python-based text mining. The source material does not announce a follow-up publication, a change in SNB communication policy or a timetable for additional research.
As an affiliate, we earn on qualifying purchases.
Key Questions
What does the SNB working paper examine?
It studies 1,063 speeches by high-ranking Swiss National Bank officials from 1928 to 2024, including their frequency, length, readability and themes.
What changes did the authors report?
The authors report that speeches became more frequent from the mid-1950s, shorter over time and more accessible in language since the 1960s. They also found that inflation gained prominence in the flexible exchange-rate era.
How did the study use ChatGPT?
The researchers benchmarked ChatGPT against traditional Python-based text-mining methods to analyze central bank communication. They describe the methods as complementary and discuss possible algorithmic pitfalls.
Does the paper show that exchange-rate changes caused the speech trends?
No such causal conclusion is established in the supplied summary. It reports that speech frequency accelerated after the transition to flexible exchange rates and that inflation became more prominent in that era, but timing alone does not prove causation.
What remains unknown from the summary?
The summary does not give detailed numerical measures for the changes or explain how closely ChatGPT’s results matched the Python-based analysis across individual tasks. Those details are in the full working paper.
Source: primary
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
