Holding to the Old Faith
Figure 1: Ten year-3 month Treasury yield spread (bold dark blue), and ten year-two year Treasury yield spread (bold dark red), and projections at current pace using 2017M01-18M08 sample (light blue and pink lines), in percentage points. August 2018 observation through August 27th. NBER defined recession dates shaded gray. Light orange denotes Trump administration. Source: Federal Reserve Board via FRED, Bloomberg, NBER, author’s calculations.
While the August ten year-three month (ten year-two year) spread is 0.87% (0.26), the corresponding figures as of 8/27 are 0.75%(0.18%), reflecting the downward trend. At the current pace the ten year-two year will invert in about 5 months.
In Chinn and Kucko (2015), we document the predictive power of these spreads for recessions. In today’s SF Fed’s Economic Letter
, Michael Bauer and Thomas Mertens follow up on their previous study, and conclude:
The three long-term spreads and the one short-term spread all have very similar predictive accuracy, with AUCs of 0.85 to 0.89. The traditional 10y–3m spread performs better than the other three spreads by a slight margin. We also assess our predictors over a sample period that excludes the so-called zero-lower-bound period, when short-term interest rates were essentially zero (2009–2015) to give the best possible chance to the short-term spread, as did Engstrom and Sharpe. This helps its relative performance and puts it about on par with the 10y–3m spread. Engstrom and Sharpe found that their short-term spread statistically dominated the 10y–2y spread, and our findings are consistent with this result. However, we also show that the traditional 10y–3m spread is the most reliable predictor, and we do not find any evidence that would support discarding this long-standing benchmark as a measure of the shape of the yield curve. It is worth emphasizing again, however, that all of these term spreads are fairly accurate predictors and quite informative about future recession risk; the differences in forecasting accuracy are small.
