Abstract No. 300964
Revision, submitted 19-April-2004

Abstract Type:	Contributed
Sponsor:	General Methodology
Title:		MARKOV MODELS FOR FINANCIAL TIME SERIES

Abstract:	                     
 
In the classical geometric Brownian motion model for stock prices the 
difference of the logarithm of the price series (the continuous rate of return, ROR) 
is a sequence of independent identically distributed Normal random variables.  
Problems in applying this model are (1) successive values of ROR may not be independent, and 
(2) the ROR may not be Normally distributed.  
The Markov model (MM) used here postulates states and transitions between them.   
This MM (1) can model dependence, and (2) the states can have different distributions, 
e.g., different Normal distributions.  
Both visible-state and hidden-state MMs are treated.  
In the hidden-state model the E-M algorithm estimates both the states 
and the distributional parameters.  
A run of one state can be considered a segment.  
Segmenting macroeconomic series by MMs can provide objective definitions of phases of the 
economy: recession, recovery, and expansion.  
Segmenting stock indices can provide objective definitions of bull and bear markets.  
Fitting MMs to stock prices can provide alternative scenarios 
for portfolio optimization, trading strategy, and options pricing.