According to some accounts, a recession appears to be imminent (10/14 60% of Bloomberg panels63% in WSJ’s October survey, 100% in Yellow/winger model).Wapo “As recession fears mount, Washington begins to weigh how to respond”. What do some models say about recession and growth?
figure 1: Recession probabilities for Treasury spreads from 10 years to 3 months (blue), and spreads from 10 years to 2 years (tan) based on a probabilistic model for the next 12 months. The 33% threshold is red. The NBER uses shades of grey to define the peak and trough dates of the recession. Source: Author’s calculations, NBER.
The 33% threshold means the 10-year to 3-month spread won’t miss any positives, but the 10-year to 2-year spread will miss the last two recessions. Only a 10- to 2-year spread would imply a recession in 12 months.
These probabilities are based on term spreads (data for October spreads up to 10/20):
figure 2: 10-year to 3-month Treasury spreads (blue), 10-year to 2-year spreads (tan), all expressed as a percentage. The October observation is related to the 10/20 data. The NBER uses shades of grey to define the peak and trough dates of the recession. Source: U.S. Department of the Treasury, NBER.
The Probit model estimated during the “Great Moderation” that the 10-year to March and 10-year to 2-year spreads yielded McFadden R2s of 27% and 18%, respectively.
In addition to recessions, one can also look at how to predict economic growth.Now, negative industrial production growth does not necessarily equate to recession (see discuss how a recession differs from a negative output movement), but they are related.
Chin-Cuco (2015) It is estimated that for every one percentage point increase in the 10- to 3-month spread over the 1998-2013 period, industrial production growth will accelerate by 1.13%.
In a recent paper, Arthur Stalla-Bourdillon, Nicolas Chatelais and I Estimate the relationship between term spreads, dividend yields and lagged growth rates for the next 12-month growth rate and compare to the dividend yield factor by industry and discover how our preferred categorization factor outperforms expectations Outperforms Competitors – Sample RMSFE Comparison. (Although outperformance relative to term spreads is not statistically significant).
Here is a picture of the current 10-year to 3-month term distribution and 12-month IP growth, one year ahead:
image 3: 10-year to 3-month Treasury maturity spread, % (blue) and 12-month IP growth, one year ahead, % (red). Observations for October are data as of 10/20. The NBER uses shades of grey to define the peak and trough dates of the recession. Sources: Treasury, Federal Reserve via FRED, NBER.
Using the model described in Chatelais, Stalla-Bourdillon, and Chinn (2022), we obtain the following IP growth estimates through September 2023.
resource: Model-Based Computing Chatelais, Stalla-Bourdillon, Chinn (2022).
Our factor model (as is the term spread and lagged IP model) predicts a modest decline in industrial production next year. During periods of negative output growth, our model performed particularly well compared to its competitors. The positive view of the total dividend yield is a bit surprising, but it is arguably true that our model should outperform when the investor discount rate (composed of the risk-free rate and the equity risk premium) fluctuates wildly.
So recession probability is rising (other models would suggest a higher probability), but our model predicts a modest decline in output (as measured by industrial production).