The likelihood of a recession hitting remains high based on our analysis of February 2020 Treasury data. Other key indicators are mixed – employment and consumer sentiment remain relatively strong while manufacturing appears weaker. Concern about COVID-19’s impact on the global economy is also having an impact on the outlook of the economy, causing chaos in the stock market and pushing down Treasury yields. Central banks have responded by cutting rates.

Our monthly U.S. recession predictor, based on a Federal Reserve model, has identified all recessions since 1955. In 2007, we used it to warn clients of the high likelihood of an impending recession.

The model predicts the probability of a recession starting 8 months out. The interactive graph below shows this probability over time, with actual recessions highlighted.

The data series used in the model are:

  1. NBER’s recession time series (1 if recession, 0 if no recession)
  2. The 10-year Treasury bond yield (Fed data series GS10)
  3. The 3-month Treasury bill yield (Fed data series TB3MS)

Estrella (2005) gives an easy-to-read overview of how the method works. Bernanke (1990) pioneered its modern use.

No predictive method is perfect. However, the term spread method is statistically robust and has given accurate signals so far (1953-2016, but with two false positives) and is viewed as a trusted indicator by experts.

  1. The latest curve from the Fed, identical to ours except that it looks 1 year out, is found here.
  2. The careful reader may notice that the curve’s prediction for past periods is based on hindsight (since the entire time series is used up till today). Properly, each period should only use data up till that period. We ran such a model and it gives the same results.