View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

short essays and reports on the economic issues of the day
2008 ■ Number 8

No Volatility, No Forecasting Power for the Term Spread
Massimo Guidolin and Allison K. Rodean


he recent disassociation between the term spread and
the real growth rate can be explained in part by the
finance fundamentals behind the concept of the term
spread. The term spread, commonly defined as the difference
between the yields on 10-year U.S. government bonds and 3month Treasury bills, can be interpreted as a risk premium:
the additional amount of compensation required to commit
wealth into long-term investments in the face of unanticipated inflation shocks. The reasoning is simple: Should inflation
accelerate unexpectedly, nominal interest rates are likely to
promptly increase. While an investor buying T-bills could
access higher yields quickly, an investor who has purchased
bonds could not and will suffer a loss since the new nominal
interest rates are higher than the bond’s rate. Because this loss
may be substantial if the bond is far from maturity, bond
investors require a reward in the form of a term spread.1
We can quantify inflation risk by using the volatility of
long-term bond yields. If there is substantial inflation risk,
investors will tend to revise often their expectations of future
inflation rates. This process affects long-term bond prices and
causes volatility in their yields.
If inflationary risks decline independently of business cycle
conditions, long-term bond yields become less volatile and the
quantity of risk declines; this causes the term spread to decrease
as well. The table shows this has been true over the past 26
years. In the Volcker era, the average realized (computed using
daily yields), annualized volatility on 10-year government
bonds had been 3.7 percent with a 2.2 percent term spread.
Under Greenspan, the average realized volatility declines to
1.3 percent and the term spread declines to 1.7 percent.
However, the past two years have been characterized by
a very low volatility in the bond market (0.7 percent) and
virtually no term spread (0.04 percent). And the past decade
has been characterized by less than half the risk formerly in
the bond market, and at the same time the term spread has
declined by approximately one-half.
Low or negative term spreads are conventionally seen as
harbingers of recession. Between January 2005 and December
2007, the term spread declined to an average of 0.35 percent
per year. Surprisingly, over the same period, real GDP maintained a brisk pace of almost 3 percent per year. Is it surprising

the U.S. term spread has stopped forecasting real economic
growth? No, because the term spread is a risk premium compensation that declines when risk disappears. For support, we
have computed correlations between the term spread and the
real GDP growth rate for two subsamples: 1981-94 and 19952007.2 We find a high and statistically significant correlation
in the first, high-volatility subsample, 0.36; with sufficient risk,
the term spread is positive and varies to reflect anticipations
of future business cycles. Over the second, low-volatility subsample the term spread has lost any association to real economic growth: The correlation is essentially zero (–0.03) and
is not statistically significant. If the term spread mostly depends
on inflationary risk and such risks disappear over time, poor
forecasting performance is expected. It is ironic that the very
success of Chairmen Greenspan and Bernanke at fighting inflation and anchoring inflationary expectations may have led to
a new era in which forecasters and policymakers struggle with
the loss of the term spread’s predictive power. The business cycle
remains difficult to forecast, although with stable inflation, the
loss of a forecasting instrument seems a small price to pay. ■

The term spread also depends on other factors, e.g., the differential liquidity of
the short and long term segments of the bond market and, in principle, the
uncertainty of future real interest rates. Even though we recognize that these factors are like to be priced in the term spread, we simply assume a direct relationship
between the term spread and the variance of excess long-term bond returns.


Consistently with the literature, we apply a lag of 3 quarters in the relationship,
i.e., real GDP growth today is predicted by the term spread 3 quarters ago. There
has been some recent debate on whether and why the term spread may actual
forecast business cycle conditions: See Anderson, R. “Yield Curve Inversions and
Cyclical Peaks.” Federal Reserve Bank of St. Louis Monetary Trends, May 2006.

term spread


Volcker era




Greenspan era




Bernanke era












Date range

Views expressed do not necessarily reflect official positions of the Federal Reserve System.

Number of