While the current economic situation can be called “unique,” Federal Reserve Bank of San Francisco researchers believe the yield curve inverting would signal an upcoming slowdown.
“Forecasting future economic developments is a tricky business, but the term spread has a strikingly accurate record for forecasting recessions,” Michael D. Bauer and Thomas M. Mertens, San Francisco Fed Economic Research Department research advisors write in an
And although, “the current environment appears unique compared with recent economic history, statistical evidence suggests that the signal in the term spread is not diminished,” they write. “These findings indicate concerns about the scenario of an inverting yield curve. Any forecasts that include such a scenario as the most likely outcome carry the risk that an economic slowdown might follow soon thereafter.”
During the past six decades, every U.S. recession was preceded by an inverted yield curve, and every inverted yield curve, except one, resulted in a recession.
In the current situation, with low interest rates, some argue “increases in the short-term policy rate may slow down the economy less than usual” and rather than recession, may be a sign of a “new normal for interest rates.”
“While these hypotheses have some intuitive appeal, our analysis shows that they are not substantiated by a statistical analysis that incorporates the suggested factors into the type of predictive models we use,” Bauer and Mertens write. “For example, including both a short-term and long-term interest rate in such models — and thereby allowing the level of interest rates to have a separate effect from that of the term spread — shows that only the difference between these interest rates, the term spread, matters for recession predictions. Similarly, various models that include estimates of the natural level of interest rates do not improve upon the predictive accuracy of the simple model with only the term spread.”
The researchers conclude “ term spread is by far the most reliable predictor of recessions, and its predictive power is largely unaffected by including additional variables.”
