An interesting article talks about the yield curve as a predictor of recessions. For various reasons (Wikipedia seems thorough on the subject) long term interest rates are usually higher than short term ones, except near the onset of recessions.
The NY Fed devotes a page to the yield-curve as a leading indicator, complete with a monthly data set back to January 1959, and charts. Indeed the correlation is impressive to the eye. If the indicator is yield curve crosses below zero, then that data shows a false negative in 1960, a false positive in 1968, and seven true positives with no false predictions since then. Other ways to use it are linked from the Fed page.
The 10-yr vs. 3-mo spread has dropped by almost half in 18 months, but is still well above zero by historical standards. Are we safe from a recession then? As they say, past performance is no guarantee of future results. The zero lower bound on rates, which the short term is close to, means that the curve will not go (more than trivially) negative, independently of any possible recession. No theory accounts for this, and so the theory may be broken here if we rely on it to feel safe from recession. OTOH, the recent drop in long term rates is not “organic”, but manipulated by the Fed’s Operation Twist, so even the dropping rates are not necessarily a danger signal.
The Fed goes further than just the spread, but derives a “probability of recession” signal from it. Zero for a positive infinite spread, one for a negative infinite spread, and in between is a (reversed) cumulative gaussian distribution with a mean of -53 basis points and a “standard deviation” of -.63. This whole enterprise seems so bogus that I am not even interested in the details. But the graph shows it to be a lagging indicator more than a leading one.
Conclusions:
- Don’t rely on the yield curve to follow its usual statistics at the zero lower bound
- The “probability of recession” from the NY Fed is a noise calculation, manufactured to impress the noise traders.