Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/259825 
Year of Publication: 
1998
Series/Report no.: 
Working Paper No. 1998:6
Publisher: 
Lund University, School of Economics and Management, Department of Economics, Lund
Abstract: 
Numerous empirical studies have shown evidence of nonlinearities in financial time series, which can be of both a deterministic and a stochastic nature. Chaos is an example of the former, and heteroscedasticity in the conditional variance an example of the latter. We apply a test, the BDS test, to Swedish Stock Index returns and detect large deviations from the IID-hypothesis. There is no evidence of chaos, and most of the nonlinearities are due to conditionally heteroscedastic error terms. We look at monthly, daily, and 15-minute return series, and find no sensitivity in the results to choice of sampling frequency. Different GARCH models often seem to explain the nonlinearities detected by the BDS test, which is particularly the case for GARCH models with t-distributed errors fitted to monthly and daily returns.
Subjects: 
BDS test
neural networks
heteroscedasticity
deterministic systems
JEL: 
C22
Document Type: 
Working Paper

Files in This Item:
File
Size
249.96 kB





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.