If no recession is coming, what can we conclude, given that most futures spread models signaled a “safe bet”? Improbable result (it’s a probabilistic world!), breakdown of historical correlations, omitted variable problem? To shed light on this question, I examine probability estimates from (i) the vanilla spread, (ii) the debt service ratio and the augmented foreign forward spread, and (iii) the spread specifications adjusted according to the term premium.
Figure 1: Estimated 12-month recession probabilities based on spread and short rate (blue), spread, short rate, debt service ratio and foreign forward spread (tan), and forward spread adjusted for term premium (green). The NBER has defined the peak to trough dates of the recession in gray. Source: NBER and author’s calculations.
Estimated 12-month recession probabilities are obtained using probit models. The first specification (blue line) is a simple forward gap model estimated over the period 1990-2023M04, assuming no recession in the United States until 2024M04. The second (tan line) is the one described in Chinn and Laurent (2024), abandoning the financial conditions index which did not provide much additional predictive power. The third (green) specification uses only a term spread, where the long rate is adjusted to remove the term premium estimated by Kim and Wright (1995). Some motivation for this modification is here.
In particular, the simple forward spread model provided the highest estimated probability of recession. The other two specifications do not give estimates exceeding 50%. It is interesting to note that the maximum probabilities are in May 2024. As we do not have statistics for May — apart from the Lewis-Mertens-Stock/NY Fed Weekly Economic Index and the Weekly economic indicators from Baumeister-Leiva-Leon-Sims (both include data for releases through May 11), it’s possible we just haven’t seen the data yet. (I still remember that in April/May 2001, many thought we had avoided a recession, based on the data available at the time).
