Last week I puzzled over the response of financial markets to the U.S. election. Since the election, the S&P 500 is now up 3%, the dollar is up 4.6% against the euro, and most remarkable of all, the 10-year Treasury rate has gone up 50 basis points. Here I offer some further thoughts on the last development.
In normal times you would expect to earn a higher return from long-term bonds compared to short-term as an added compensation for being exposed to various risks. Over the last half century, the yield on 10-year Treasury bonds averaged 1.56 percentage points (or 156 basis points) higher than 3-month Treasury bills. But while the 10-year rate is up 50 basis points since the election, the 3-month rate has only risen 4 basis points. What we’ve seen is a rapid increase in the spread between long and short rates and a sharp steepening of the yield curve.
Yield on 10-year Treasury bonds minus that on 3-month Treasury bills, monthly June 1961 to Nov 2016. Blue line corresponds to average historical spread of 156 basis points.
In some historical episodes, short-term securities have a higher yield than long-term and the yield curve becomes inverted. These brief episodes of a negative spread usually come near the end of an episode of tightening short-term rates. If investors anticipate falling interest rates, an anticipated capital gain on a long-term bond can help compensate for the fact that its current yield is less than that on a 3-month T-bill.
Top panel: Spread between 10-year and 3-month yields. Bottom panel: 3-month yield.
In principle it should be possible to calculate how much of the movement in the spread represents a changing assessment of future short-term rates and how much is a change in the premium earned for holding long-term bonds. If we have a model for forecasting short-term rates, we could calculate how much the spread should change due to the first factor, and attribute the difference between the actual spread and the predicted spread to changes in the term premium. Of course, the answer one arrives at can be quite sensitive to the model usedfor forecasting. Below is a graph of one popular estimate of the term premium.
