THE BOND MARKETS: Estimation 4


Therefore, we rerun the estimation carried out in Table 4 allowing for a non-linear drift term. The stochastic process specification is as follows:
w6631-20
w6631-21

the mean-reversion introduced by the jump process is not sufficient to rule out the non-linearity in the drift term. Click Here One may conclude that while jump processes may ameliorate the non-linearity in the drift, it is still a feature that appears robust to enhanced specifications such as that introduced in this paper.

Estimation using the Method of Moments. The method of moments has the advantage that the jump distribution can be modelled quite generally, and the estimation scheme is easy to implement. Empirical estimation for the method of moments is undertaken using the standard Hansen [28] efficient generalized method of moments estimation approach. We estimated two models: (i) a pure diffusion model and (ii) the jump-diffusion model.

First, the pure diffusion model was estimated. All first four moments of the distribution were used for the estimation so as to be consistent with the jump-diffusion model. As in the paper by Chan, Karolyi, Longstaff and Sanders [18]. the instruments used here are a constant, and the lagged value of the short rate. For the pure diffusion model, the moments used are a special case of the moments in Section 2.3, where the jump variables are eliminated from the moment expressions

Table 12. Method of Moments Estimation
w6631-22