Abstract of a new paper co-authored with Laurent Ferrara, “The Missing Recession: Revisiting the Predictive Power of Term Spreads”:
The failure of a recession to occur in 2024 has forced a reappraisal of the term as the most reliable recession predictor for the US. In this paper, we re-evaluate the predictive power of the term spread, and explain false positive signals in terms of sample specificity, simultaneity, and omitted variables, focusing on the United States, but including other developed economies where the term spread has proven to be a useful predictor in the past. We confirm that before the 2020 recession, the yield curve was an accurate predictor of recession in the US, Canada, and Germany, but proved to be less accurate. For the US, accounting for term premia does not significantly improve predictions, when considering future unemployment expectations. More definitively, in line with Borio, et al. (2020), we find that including the debt service ratio approves the failure of a recession to emerge in 2024 after a strong yield curve inversion observed at the end of 2022, for the US and Germany.
Focusing on the United States, it is clear that most measures of the term spread (10yr-3mo, 10yr-2yr, etc.) indicate a high probability of recession in 2023-24, due to a large and persistent inversion in 2022. Relying on pre-2020 datathe possibility of a recession looks like this.
Figure 1: Top panel – spread US 10yr-3mo,%; bottom panel – recession probability, using regression estimates up to 2019M12. NBER designated peak-to-trough recession dates are shaded gray.
There are many explanations for the term spread mis-fire 2024: the process of generating data has changed since 2020, the behavior of the premium term (which is part of the term spread) has changed, the omission of Taylor Rule variables like the expected unemployment rate or the omission of financial variables like the financial cycle has biased the forecast. It is also possible that “by chance” a recession simply does not occur; after all, one is never 100% confident.
In this paper, we examine these possibilities, either by estimating the specification of the term spread over the full sample, or adding it to the key variable.
Source: Table 2 from Chinn and Ferrara (2026).
Note that the proportion of variance explained using the spread term is higher in the pre-2020 sample, and lower in full (0.33 vs. 0.26). Add in separate assessment terms premium (est. ACM from NY Fed) does not replace full sample results. However, adding the expected change in the unemployment rate (because the Taylor rule indicates that the Fed funds expected in the future will be lower if the market expects a slowdown/recession) does not result in a significant coefficient for this variable (even pseudo R2 fundamentally unchanged).
Only including a proxy for the financial cycle – the debt service ratio (DSR) – raises the proportion of well-explained variance, to 0.57. It also turned out to be the most “fit” – in the sense that it is not a false positive, when holding the recessions of 1990-91, 2008-09, this is the equation combining DSR (teal line in the figure below, marked & DSR).
Source: Figure 5 from Chinn and Ferrara (2026). NBER designated peak-to-trough recession dates are shaded gray.
The finding that the inclusion of DSR helps predict recessions accurately, although some limit the forecast period to before 2020. Then the 1990-91, 2001, and 2008-09 recessions are predicted, but not the 2020 pandemic (or the 2024 recession).
It is not accounting for Taylor’s rule is useless (column 4 in Table 2). Including the expected change in the unemployment rate (conjoined with the use of the Fed funds rate instead of the 3 month rate) reduces the estimated probability of a recession for 2024 below the threshold of 40% (blue line, marked &UE), while holding the previous recession, save 2008-09 one. So in that sense, it is not clearly considered that the Taylor rule can be part of the explanation for the recession call for 2024.
Read the entire paper here, which also examines the Canadian and German experiences. This paper is a revised version of the paper described here):