Séminaire OrléansUnconditional time-dependent variance as the source of financial volatility
Michael Levine (Purdue)
Thursday 24 May 2012 14:00 - Orléans - Salle de Séminaire
Every researcher who ever had to deal with financial time series would quickly name two ways of modeling conditional heteroscedasticity: 1) using an ARCH/GARCH framework and 2) employing stochastic volatility models. However, both of these approaches are, effectively, about modeling the conditional volatility; there is certain evidence in the literature that, if there is some volatility due to unconditional time-dependent variance, GARCH type models fail. In this talk, I will describe modeling a time-dependent volatility by means of the smooth function of time that DOES NOT depend on the past values of the process. I will describe implications of that assumption, place the resulting model in the econometric literature context, and establish some of its properties. In particular, I will show that such a model can be easily estimated and asymptotic distributions of its parameters can be established without much trouble. I will finish the talk with some simulations illustrating good empirical behavior of the proposed estimation procedures.