theoretical spread. The threshold model allows reducing the uncertainty associated to the volatility
of term premia. The augmented predictive power of the spread in both regimes follows directly
from the lower level of uncertainty that characterizes each regime. This story is also consistent with
the idea put forward by Mankiw and Miron (1986), who suggest the high interest rate predictability
leads to small downward bias of the slope estimates. In particular, they attribute to the random walk
behaviour of the short term rate, due to the interest rate smoothing policy adopted by the Federal
Reserve, has affected the empirical support for the EH.
. | |||||||
_________________________Standard Deviations_________________________ | |||||||
____________________long term maturity ( n )____________________ | |||||||
6 |
12 |
24 |
36 |
60 |
120* | ||
reg 1 |
spread |
0.1818 |
0.4134 |
0.6880 |
0.8940 |
1.0456 |
1.1649 |
thspr |
0.3500 |
0.6723 |
1.0154 |
1.2508 |
1.6397 |
1.6763 | |
tp_______ |
0.3351 |
0.6394 |
0.9227 |
1.1256 |
1.2866 |
0.8816 | |
reg 2 |
spread |
0.3238 |
0.5074 |
0.7689 |
0.9628 |
1.2448 |
1.4816 |
thspr |
0.6053 |
0.7307 |
1.0909 |
1.2906 |
1.7443 |
2.0250 | |
_JE_____ |
0.6425 |
0.6663 |
0.8824 |
0.9863 |
1.1154 |
1.3404 | |
linear |
spread |
0.2349 |
0.4656 |
0.7447 |
0.9395 |
1.1693 |
1.4165 |
thspr |
0.5268 |
0.9174 |
1.4810 |
1.8339 |
2.1609 |
2.5069 | |
tp_______ |
0.5699 |
0.9578 |
1.5253 |
1.8453 |
2.0577 |
2.4383 | |
Table 9
Thornton (2003) points out that the conventional test of the expectations hypothesis tends to
generate large estimates of the slope coefficient depending on the relative variance of the short term
to the long term rate, suggesting that the uncertainty is largely connected to the conduct of the
Federal Reserve. In particular, the more volatile short rates are relative to long rates, the closer the
estimated β to one. The uncertainty affecting the economy, in our model captured by the level of
term premia, influences the empirical validation of the expectations hypothesis. The rationale works
as follows. Suppose that an exogenous unanticipated inflationary shock hits the economy generating
a massive response by long rates; unexpected important variations in long rates increases volatility,
which in turn bias the slope estimate downward. In Table 10 for any combination of maturities (n,
3) we show the values of the ratio between the variances of the short and the long rates. The bottom
panel refers to the single regime (January 1964 - June 2002). In the top panels we report the
variance ratio for the threshold models, both when the threshold is the term premium, and when the
threshold is the absolute value of the term premium. The relative variance increases with maturity n
both in the entire sample and in each regime determined by the level of the term premium; whereas,
in regimes split by the absolute value of the term premium the relative variance is increasing with
maturity only below the threshold (regime one, that is when the term premium is low in absolute
25