Quantifying the Beige Book’s “Soft” Data

The Federal Reserve’s Beige Book is a narrative published eight times a year, typically two weeks prior to a Federal Open Market Committee (FOMC) meeting. It summarizes regional changes in economic activity since the previous FOMC meeting based on comments from business and community contacts in each Federal Reserve district. Some economists believe that the Beige Book’s “soft” content is inferior to hard economic data (PDF), such as the monthly Employment Situation report. We show that a very basic approach to text analysis reveals the Beige Book’s soft content is actually quite firm.For a complete analysis, see our November 2024 Federal Reserve Bank of St. Louis Working Paper, “The Beige Book and the Business Cycle: Using Beige Book Anecdotes to Construct Recession Probabilities.”

Beige Book Sentiment and GDP Growth Move in Tandem

The figure below plots a simple measure of economic sentiment as reflected in Beige Book reports and year-over-year growth of real gross domestic product (GDP), with recessions shaded in gray. To calculate sentiment for a given Beige Book, we take the difference between the number of positive words and the number of negative words that appear in the report and divide that difference by the Beige Book’s total number of words.The lists of positive and negative words come from Elaine Henry’s October 2008 Journal of Business Communication article, “Are Investors Influenced by How Earnings Press Releases Are Written?”; Tim Loughran and Bill McDonald’s February 2011 Journal of Finance article, “When Is a Liability Not a Liability? Textual Analysis, Dictionaries, and 10-Ks”; and Luca Barbaglia, Sergio Consoli, Sebastiano Manzan, Luca Tiozzo Pezzoli and Elisa Tosetti’s May 2022 working paper, “Sentiment Analysis of Economic Text: A Lexicon-Based Approach.”

Beige Book Sentiment and Year-over-Year Real GDP Growth

SOURCES: National Bureau of Economic Research, Bureau of Economic Analysis and authors’ calculations.

NOTES: Data are quarterly. GDP growth and Beige Book sentiment have been standardized with respect to their averages and standard deviations from the second quarter of 1970 to the third quarter of 2024. As a result, the positive (negative) values shown here indicate that GDP growth and sentiment are above (below) average, rather than that they are positive (negative). Quarters that experienced recession as defined by the National Bureau of Economic Research are shaded in gray.

Both economic sentiment and real GDP growth are shown with respect to their long-term averages. Generally, the two series track each other quite closely. The spread between economic sentiment and GDP growth was considerably tighter in the decades prior to the 2008 global financial crisis, though even after 2008 they have tended to move in tandem. As of the November 2024 Beige Book, sentiment was approximately 0.3 standard deviations below its historical average, while the third estimate of real GDP for the third quarter of 2024 showed yearly growth was just in line with its historical average.

The Beige Book Highlights Major Current Events

Beige Book anecdotes also contain references to events that are not directly related to the business cycle. The shock indexes in the figure below track the occurrence of certain key words in the report text. The Physical Shock Index reflects the frequency of words like “hurricane,” “tornado,” “earthquake,” and others related to natural or environmental disasters. The Political Shock Index similarly tracks words like “election,” “war,” “strike,” and others.These shock indexes are not sentiment indexes; they track only the number of times that “shock” words appear in the report relative to the total number of words in the report. These indexes are similar to the inflation and labor market indexes constructed by Charles Gascon and Devin Werner in their June 2022 Economic Synopses article, “Does the Beige Book Reflect U.S. Employment and Inflation Trends?”

In the November 2024 Beige Book, both indexes remained elevated and contributed to negative economic sentiment, though they had come down from October’s Beige Book. Hurricane Helene made landfall in late September, and the October reports from the Richmond and Atlanta Feds—whose districts include the states most severely affected by the storm—contained some references to its impact. Contacts in the Richmond Fed’s district continued to note damage and disruptions to manufacturing, tourism and real estate in the November report. The increase in the Political Shock Index was driven by references to both uncertainty in anticipation of the November elections and the dockworkers’ strike.

Charting References in the Beige Book to Physical and Political Shocks

SOURCES: National Bureau of Economic Research and authors’ calculations.

NOTES: The shock indexes show the percentage of words in the Beige Book that appear on lists of shock words reflecting natural disasters and political/social events. The lists of shock words were created by the authors. Data are monthly from April 1970 to November 2024. Months that experienced recession as defined by the National Bureau of Economic Research are shaded in gray.

The Beige Book and Business Cycle Turning Points

We use a probit model to track our measure of economic sentiment and the shock indexes obtained from the Beige Book, which can be mapped quite accurately to recessionary periods.A probit regression is commonly used to find the probability of an outcome given a set of independent variables. The probit model here uses only the current Beige Book sentiment score, a one-period lag of the sentiment score and the current shock indexes to determine the probability that the National Bureau of Economic Research will consider the current period a recession. The real-time probability of a current-quarter recession derived from the Beige Book closely matches the probability of a decline in current-quarter real GDP reported by the Philadelphia Fed’s Survey of Professional Forecasters, as shown in the figure below. The Beige Book recession probabilities turn upward sharply around the time of a business cycle peak and downward sharply around the time of a trough.

Tracking the Probability of a Recession by Forecasters and Beige Book Sentiment

SOURCES: National Bureau of Economic Research, Survey of Professional Forecasters and authors’ calculations.

NOTES: Data are quarterly from the first quarter of 1980 to the fourth quarter of 2024. Quarters that experienced recession as defined by the National Bureau of Economic Research are shaded in gray.

The past 10 quarters—the period since the height of the COVID-19 pandemic—have been unique in that Beige Book recession probabilities have, mostly, remained above 10%. However, as of the November 2024 Beige Book, the probability that there was a recession in the fourth quarter of 2024 was only 4%. Past recessions have not occurred until the probability stays above 47% for two consecutive reports. It’s also worth noting that the model we use doesn’t predict recessions far in advance: Typically, the probability of recession increases by, at most, only 4 or 5 percentage points in the quarter prior to a downturn, and there have been several such increases unaccompanied by recessions.However, in a similar exercise, the Beige Book was shown to have some ability to forecast recessions. See Ilias Filippou, Christian Garciga, James Mitchell and My T. Nguyen’s April 2024 Federal Reserve Bank of Cleveland Economic Commentary, “Regional Economic Sentiment: Constructing Quantitative Estimates from the Beige Book and Testing Their Ability to Forecast Recessions.”

Conclusion

The economic sentiment score and associated recession probabilities can help analysts quantify the Beige Book and track broad trends within the economy over time. However, the value of the report comes from the nuance of the granular anecdotes that Federal Reserve economists and officials collect from business and community leaders.

Neglecting to read the Beige Book and instead focusing only on the numerical sentiment scores derived from it may cause an observer to detect big fluctuations in economic conditions without understanding the driving forces behind those changes. Accordingly, the Beige Book sentiment scores and recession probabilities that we discuss should be seen as a tool to complement the written report and other measures typically used in economic analysis.

Notes

1. For a complete analysis, see our November 2024 Federal Reserve Bank of St. Louis Working Paper, “The Beige Book and the Business Cycle: Using Beige Book Anecdotes to Construct Recession Probabilities.”
2. The lists of positive and negative words come from Elaine Henry’s October 2008 Journal of Business Communication article, “Are Investors Influenced by How Earnings Press Releases Are Written?”; Tim Loughran and Bill McDonald’s February 2011 Journal of Finance article, “When Is a Liability Not a Liability? Textual Analysis, Dictionaries, and 10-Ks”; and Luca Barbaglia, Sergio Consoli, Sebastiano Manzan, Luca Tiozzo Pezzoli and Elisa Tosetti’s May 2022 working paper, “Sentiment Analysis of Economic Text: A Lexicon-Based Approach.”
3. These shock indexes are not sentiment indexes; they track only the number of times that “shock” words appear in the report relative to the total number of words in the report. These indexes are similar to the inflation and labor market indexes constructed by Charles Gascon and Devin Werner in their June 2022 Economic Synopses article, “Does the Beige Book Reflect U.S. Employment and Inflation Trends?”
4. A probit regression is commonly used to find the probability of an outcome given a set of independent variables. The probit model here uses only the current Beige Book sentiment score, a one-period lag of the sentiment score and the current shock indexes to determine the probability that the National Bureau of Economic Research will consider the current period a recession.
5. However, in a similar exercise, the Beige Book was shown to have some ability to forecast recessions. See Ilias Filippou, Christian Garciga, James Mitchell and My T. Nguyen’s April 2024 Federal Reserve Bank of Cleveland Economic Commentary, “Regional Economic Sentiment: Constructing Quantitative Estimates from the Beige Book and Testing Their Ability to Forecast Recessions.”