The Debt-Service Ratio and Estimated Recession Likelihood

Or, why I nonetheless assume a recession is feasible at 2024H1.

The BIS has simply launched debt-service ratios for 2023Q1. Utilizing this variable, and time period unfold and overseas time period unfold, the estimated chance of a recession in 2024Q1 has risen considerably, though nonetheless far beneath the corresponding determine utilizing solely a time period unfold/quick fee.

Determine 1: Estimated chance of NBER-defined recession utilizing 10yr-3mo time period unfold plus 3 month Treasury fee (blue), and specification augmented with overseas time period unfold and debt-service ratio (tan). Debt-service ratio is for personal nonfinancial sector. NBER outlined peak-to-trough recession dates shaded grey. Supply: Creator’s calculations and NBER.

The rise in estimated chance for being in a recession month in rose from 12% in 2024M02 to 41% in 2024M03. This highlights the truth that the probit regression includes estimating a nonlinear relationship between the dependent and unbiased variables. The rise in debt-service ratio seems fairly modest.

That being mentioned, the augmented specification implies a chance far beneath that from the term-spread and quick fee solely specification (in March 2024, it’s 68%, in comparison with 41% from the augmented specification). For motivation to be used of debt-service ratio as a proxy for the monetary cycle a la Borio-Drehmann-Xia, and the overseas time period unfold a la Ahmed-Chinn, see this post. Extra cross-country evaluation in work with Laurent Ferrara in slide deck here.

Determine 2 depicts the evolution of the time period unfold and the debt-service ratio for the non-public nonfinancial sector.

Determine 2: 10yr-3mo Treasury unfold, % (blue, left scale), and debt-service ratio, % (tan, proper scale). NBER outlined peak-to-trough recession dates shaded grey. Supply: Treasury by way of FRED, BIS, NBER, and writer’s calculations.

Observe that the time period unfold went extra unfavorable in 2023Q1 because the debt-service ratio rose. The rise in estimated chance within the un-augmented specification was 18 share factors, and 29 share factors for the augmented.

One may ask why one ought to want the augmented specification over the the unaugmented. The pseudo-R2 for the time period unfold plus quick fee specification (1985-2023M08) is 0.28, whereas that for the debt-service and overseas time period unfold is 0.60. (Debt-service alone has a pseudo-R2 of 0.32, however implies primarily zero chance of recession in 2024M03).

As well as, the augmented specification captures 3 out of 4 recessions through the pattern interval, utilizing a 50% threshold (lacking the 2020 recession). A time period unfold plus quick fee catches solely the 2001 recession, whereas a debt-service ratio alone specification misses all 4 (coming shut for the 2007-09 recession). Utilizing a 40% threshold reduces the discrimination between the fashions, with each the time period unfold plus quick fee and the augmented catching 3 out of 4 recessions (debt-service ratio solely catches the 1990-91 and 2007-09 recessions). All specs miss the 2020 recession.

An attention-grabbing apart: Even in the event you assume the time period unfold is not a dependable predictor (e.g., Putnam), that doesn’t imply you may loosen up.

So, whereas the “tender touchdown” view is gaining dominance, the above outcomes are in keeping with the view of a downturn in 2024H1, which garners about 53% within the FT-Sales space College macroeconomists survey: