In the 1th Quarter 2004 release of SPV the four standard models have been re-estimated. Any changes in the estimated parameters are due to the fact that the estimation window is moved one term forward, covering January 1999 to January 2004. This change in the sample is very small and only results in a minor change of the model parameters.
In addition to the general model update we focus on the option adjusted duration (OAD) which comes as output from the calculations of RIO. We compare the theoretical durations to the actual price sensitivities found in the data.
Inspired by the work of DeRosa et al (1993) we develop a new way of gauging the difference between theoretical and actual price sensitivities, allowing us to examine biases of the OAD. Using this test we find that the OAD is biased upwards.
By using various clinical experiments we conclude that, in the interest environment for the last years, negative convexity is most likely not the cause for the upward bias of OAD while we find evidence that non-parallel shocks, not captured with the OAD, could be the cause of the bias. Finally we also discuss that the assumption of a fixed option adjusted spreads in OAD calculations could cause the bias also. This idea is currently being pursued by the research team at ScanRate.