pcess609
Using large language models could allow economists to turn narratives into data. This is especially useful when analyzing the Federal Reserve’s Beige Book, a report released eight times a year that compiles anecdotes about the economy from all 12 Fed countries. districts and provides a national summary of the economy.
Their conclusion: the sum of the parts is more than the whole.
Four economic researchers applied that technology to determine whether the Beige Book could help identify or predict recessions. Specifically, they used FinBERT, a variant of BERT, a deep learning model using natural language processing developed by Google researchers in 2018.
The model is more useful than word counting or simpler models of language processing. “Using the surrounding text, rather than simply reading from left to right, BERT aims to establish the context of the text and thus infer the meaning of language that might otherwise be ambiguous,” the authors wrote.
“Our goal is to establish (reduced-form) national and district-level sentiment statistics in the Beige Book and examine their predictive power for U.S. recessions,” they wrote in an economic commentary published by the Cleveland Fed. “As such, our measure of sentiment is an aggregate measure, meaning it quantifies sentiment across different areas of the economy, such as overall economic activity, labor markets, prices, and different sectors.”
The authors — including two Chicago Fed economists and a visiting assistant professor at Washington University in St. Louis, and a postdoctoral research associate in finance at the same university — were particularly interested in how well overall national sentiment , as conveyed in the Beige Book, with respect to the consensus sentiment derived from the 12 districts.
They found that “national economic sentiment does not always fall in the middle of district-level estimates, as one might expect.” And as of 2021, national sentiment is more positive than in most individual districts, they said.
National sentiment, it seems, weighed more heavily on some districts than others. And their analysis indicated that the national measure underweighted the Cleveland, Minneapolis, Philadelphia and St. Louis districts.
By contrast, economists found that sentiment in Chicago, Minneapolis, Philadelphia, Richmond and San Francisco was “very useful in predicting current turning points in the US business cycle.” The data showed that as sentiment improved in each district, the chances that the US economy was in an expansionary phase increased.
And the Boston district has proven very useful in making forecasts three and six months ahead, they said.
From the mid-1980s until the start of the Covid-19 pandemic in March 2020, the authors said they observed no cases of false alarms, “that is, recession probabilities exceeding about 40% without a recession actually occurred.”
As of March 2024, the analysis suggested that “the probability of a national recession was low, conditioned by country- and district-level sentiment data from the March 2024 Beige Book. But this probability has rebounded considerably since the pandemic-induced crisis.” . early 2020 recession.”
“In summary, observed heterogeneity in economic sentiment at the district level can be used, in addition to information contained in national economic sentiment, to better predict US recessions,” they said.