 |
 |
Thanks to all who helped me with my EGARCH queries...especially Pieter and Jean-Philippe. Being relatively new to GARCH estimation, I have another question with regard to interpretation. I am estimation a GARCH(1,1) model for the pound-dollar exchange rate. There are no covariates in the conditional mean. The conditional variance, however, includes three variables that measure changes in the political and economic situation in the UK. I include a dummy variable for the 1992 currency crisis, a continuous indicator of public approval for the party in power and an interaction between these two variables. The results I get are: MEAN _constant .0048 VARIANCE _constant -10.56 crisis 1.96 approval -1.22 interaction -.020 ARCH arch .056 garch .764 All variables are individually significant. Here is my question: how does one interpret the coefficients on the variables in the VARIANCE? If this was traditional OLS, I would be comfortable saying that the currency crisis caused the conditional variance to increase by 1.96 and that an increase in government approval by one-percent decreases the conditional variance by 1.22%. I would then proceed to calculate the influence of the interaction by looking at different levels of approval. In the GARCH context, I am not sure (i) how to interpret especially since the effect is non-linear (I think) and (ii) if and how the arch and garch terms fit into the overall interpretation. Can anyone offer suggestions about either interpretation or references to the relevant literature that discusses these issues. Many thanks, David Leblang University of Colorado