The Bank of France has published an article introducing the market-implied probability of economic contraction indicator, or PICON, which uses machine learning to measure how financial markets perceive recession risk in the United States and euro area. The indicator applies a random forest model to daily data across equities, sovereign and corporate bonds, currencies, commodities and derivatives, allowing it to capture interactions that yield curve models alone may miss. It should be interpreted as a measure of market perceptions rather than a recession forecast. The model identified three of the four U.S. recessions since 1990 with probabilities above 50% and produced fewer false signals than a benchmark yield curve model during 2022-24. In the 2022-25 test period, it distinguished stress associated with economic concerns, including the European energy crisis, banking failures, prolonged high interest rates and U.S. tariff announcements, from episodes of mainly financial stress. Asset classes contribute differently across time horizons, with equities and credit providing more information in the short term and yield curve slopes becoming dominant after three or four quarters. Against the backdrop of the war in Iran, PICON placed the market-implied probability of a euro area recession within two quarters at 21% as of April 14, 2026.
2026-09-10Bank of France
Bank of France introduces machine learning indicator for daily market implied recession risk in the United States and euro area
The Bank of France has introduced PICON, a machine learning indicator that uses daily financial market data to measure perceived recession risk in the United States and euro area. It combines signals across asset classes and distinguishes macroeconomic concerns from purely financial stress, but is not a recession forecast. The indicator placed the probability of a euro area recession within two quarters at 21% as of April 14, 2026.