In this fourth release of SPV in 2002 the focus has been on the development of
a new benchmark model for MBB valuation.
The traditional model supplied with the RIO system has undergone a gradual
improvement in the last years, but with the introduction of the RIO 4 software a
number of new models and numerical tools have become
available and as a result of that we have decided to do a full respecification of the model.
The major subjects treated in this release are as follows:
- Construction of a new swap curve with more desirable properties.
- Evaluation of two specifications of the volatility structure of the extended
CIR model.
- A discussion of the possible bias in MBB-values when prepayment forecasts are used to fix future prepayments in the
MBB model.
1.
Currently the swap curve is taken directly from Reuters. The problem with this
curve is that it only contains assets with a maturity less than 10 years.
Further, since it is bootstrapped a smooth forward curve cannot be guaranteed.
A new benchmark model will rely on a volatility structure calibrated
to quoted implied volatilities of swaptions and it is evident that a smooth
forward curve is important for stable results.
In this release vi suggest a solution based on a direct estimation using
swaps and money market deposits.
We find that the proposed method solves the observed problems with the
forward curves and in general has a nicer shape than the bootstrapped curve.
2.
We examine two specifications of the volatility structure of a CIR model.
One of the specifications is quite general, allowing the model to fit
swaptions of both long and short maturity very well. The alternative
is a constant volatility specification.
These two specifications are examined with respect to the time series behavior
of parameters, deviation measures, values of specific swaptions and finally
the out-of-sample properties of the two specifications.
3.
A recurring subject of SPV has been the use of data on preliminary
redemptions (CK93) to give a partial forecast of the next prepayment rate.
In this release we examine the implications on valuation, when forecasts are used to fix prepayments, both over groups with different levels of individual prepayment and
over different future states of nature.
We find that one has to be
careful how this is done since it has nontrivial effects on the value and
sensitivity measures of the bond.
In the next release of SPV we present a range of improved techniques for the integration of individual prepayment forecasts in the pricing model.