A 16-variable 'hard data' factor model to assess 'late cycle' turning points

With the US economy seemingly late cycle (judging by the labor market) and the market already pricing a sizeable Fed hiking cycle, the focus is back on recession risk. We use different types of recession probability models (“hard data” factor models, yield curve models, credit risk models and survey based models) and this first note serves as a refresher on how the hard data factor model (summarizing 60 years of activity data) works and how to interpret the data. We will compare the recessionary signal from the real side with the spread signal in the next installment, coming shortly.

Currently no sign of recession around the corner

It turns out that the factor model we ran pre-pandemic is quite stable because it was conceived to be robust to extreme outliers and outsized generalized volatility. Running the model with or without the pandemic in the sample changes little to the estimated cycle or recession probability indicator depicted below. The upshot is that currently, the state of real activity is very far from signaling a recession—the implied probability of entering a contractionary phase is essentially zero (0.02% to be exact). Furthermore, our probability indicator usually turns several months ahead of an officially sanctioned(National Bureau of Economic Research (NBER) recession (depending on the nature of the shock that generated the recession), and it can take again several months to shoot up to that warning level. Currently with March monthly data in hand—remember, we have only real data, no PMIs or financials in the factor—there is no recession right around the corner. This means there is still leeway for policy to steer the economy to the right landing zone from the perspective of the 'real' side.

Explore related articles