Building on Ben Bernanke’s widely used recession probability model, we invented a better model using exactly the same data.
BOSTON, MA, UNITED STATES, January 22, 2026 /EINPresswire.com/ — What is the probability of a U.S. recession 9 months out? We present one number, daily updated.
Our TelluBase U.S. Recession Predictor represents a significant advancement in economic forecasting, building upon established methodologies with innovative enhancements.
This advanced model employs a P.I.D. (Proportional-Integral-Derivative) approach that captures multiple dimensions of yield curve dynamics, providing more nuanced and accurate predictions than previous versions.
By analyzing the relationship between short-term and long-term Treasury yields, the model identifies patterns that have historically preceded economic downturns. This represents a meaningful step forward in quantitative economic analysis, combining rigorous statistical methods with real-time market data to deliver actionable insights.
The method builds on the well-established use of the yield curve as an indicator of upcoming recessions.
𝗠𝗲𝘁𝗵𝗼𝗱
Prof. Ben Bernanke, later chairman of the Federal Reserve, in 1990 developed a method to find the probability of future recessions. He kept the model simple using only the term spread between long-term bills and short-term bonds and applied to a probit model with actual recessions, as determined by the National Bureau of Economic Research (NBER), as the dependent binary variable.
The beauty of the model is its simplicity. Bernanke did not try to find funky variables that would work for a while. Instead, he had a sound theoretical reason for why the term spread is suitable.
Tellusant has developed a new version of the model. We wanted to use exactly the same data, avoiding variable fishing that may work for a while, but is spurious.
The Bernanke model is proportional. It uses a P controller with 12 months lag. It means the prediction is 12 months out.
We used P and added integral I and derivative D elements (working on the same term spread data). That is, a P.I.D. controller. The periods for both I and D were found to be nine months, while P remains at 12 months. This P.I.D model has an optimum at 9 months prediction horizon (found by reviewing the differenced cross-correlation function).
Comparing the Aikake Information Criterion tells us that there is close to zero probability that the Bernanke model is better. This is also seen visually because the Tellusant model gives significantly stronger positives than the Bernanke model.
Neither the Tellusant nor Bernanke model are perfect. Both give false positives occasionally (notably in 2025; although once NBER makes its determination there may be a recession in 2025, although shallow), but so far no false negatives.
𝗙𝘂𝗿𝘁𝗵𝗲𝗿
Tellusant is a research-heavy company. Find more of our intellectual property in our Publications Repository and in our chairman’s personal site SCANBACK think.
Philip Burginyoung
Tellusant, Inc.
email us here
Visit us on social media:
LinkedIn
Other
Legal Disclaimer:
EIN Presswire provides this news content “as is” without warranty of any kind. We do not accept any responsibility or liability
for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this
article. If you have any complaints or copyright issues related to this article, kindly contact the author above.
