Valls, Pedro: Alternative Models to Extract Asset Volatility: A Comparative Study
World Conference Econometric Society, 2000, Seattle

Nuno Miguel de Almeida, University of Sao Paulo
Luiz K. Hotta, UNICAMP
Luiz Alvares Rezende de Souza, University of Sao Paulo
Pedro Valls, IBMEC Business School
Alternative Models to Extract Asset Volatility: A Comparative Study
Session: C-11-20  Tuesday 15 August 2000  by Valls, Pedro
This paper presents an empirical comparison in the estimation of the volatility of three Brazilian financial series: a Brazilian Brady bond (the Cbond), the most tradable share in Sao Paulo Stock Exchange (Telebras PN) and the Brazilian Real / US Dollar exchange rate, using different modelling methods. The models used are: XARCH family, Stochastic Volatility (SV) and the switching in the variance model (SWARCH). The comparison is done using three criteria: loss functions, which compare the square of the estimated volatility with the instantaneous volatility, a procedure proposed by Herencia et alii (1998) which used prediction confidence intervals and one-step-ahead prediction, and a prediction exercise for the last 100 observations. In general the SV model presented the best performance although it is dominated by other models in some criteria.

File created by Jurgen Doornik with eswc2000.ox on 2-01-2001