Estimating EGARCH-M models: Science or art?
Quarterly Review of Economics and Finance
This paper shows that the EGARCH-M model should be estimated with caution. Regardless of the assumption made regarding the conditional error distribution, the EGARCH-M model is sensitive to the choice of starting values and the degree of computer precision. However, a simple forecasting example shows that these issues only become a concern when the EGARCH-M model is estimated on a high-precision computer, under the assumption that the conditional errors follow a generalized error distribution. The instability of the risk aversion parameter in the mean equation is the greatest influence on these results.
Original Publication Date
DOI of published version
St. Pierre, Eileen F., "Estimating EGARCH-M models: Science or art?" (1998). Faculty Publications. 3934.