value); the pattern it is irregular across maturities above the threshold (regime two). These results
are consistent with the estimated magnitude of the slope coefficients reported in Table 6.
____________________Relative Variance____________________ | |||||||
______________long term maturity ( n )______________ | |||||||
6 |
12 |
24 |
36 |
60 |
120 | ||
tp”, m |
reg 1 reg 2 |
1.0123 0.9420 |
1.1532 0.9972 |
1.2764 1.1255 |
1.5082 1.2099 |
1.7252 1.4338 |
1.6974 1.7357 |
∣tpt'"∣ |
reg 1 |
0.9896 1.1384 |
1.1011 1.058 |
1.2284 1.1257 |
1.3662 1.0957 |
1.6106 0.9795 |
1.7886 1.7678 |
linear |
0.9766 |
1.0581 |
1.1339 |
1.2162 |
1.2954 |
1.3748 | |
Table 10
The predictability, or the unpredictability, of short term rates, as measured by the unexpected
change in the spread (the term premium) affects the empirical testing of the expectations
hypothesis. The crucial assumption of regime-dependent behaviour of the term premium seems to
explain this empirical fact. An example will clarify the underlying principle. Suppose the economy
is hit by a negative supply shock, like a sudden increase in the oil price; suppose further that the
effects on inflation are expected to be long-lasting; hence, it follows also a sharp and permanent
increase in long term rates. The yield curve becomes steeper and term premia tend to increase. The
shock affecting the interest rates dynamics also increases the probability of switching regime.
Market participants rationally forecast the transition to the new regime and, in such enriched
informative context, are able to improve interest rates forecasts. For example, forward looking risk-
averse agents know that large values of the term spread are associated with increasing short term
rates; therefore they may anticipate future monetary policy tightening. Threshold models for risk
premia provide with a technical framework that works as uncertainty reducer; the final effect is to
diminish the unpredictability of interest rates. In sum, risk-averse agents rationalize their attitude
towards risk and exploit financial information more effectively. In this sense, the main advantage of
the threshold approach is to provide the empirical framework that is consistent with completeness in
financial markets. In each state of the world agents know exactly the macroeconomic and financial
environment in which they are playing.
In the next Section we provide evidence that non linear modelling is also suitable to analyse the
informative content of the term structure for predicting future inflation.
26