Please use this identifier to cite or link to this item: http://hdl.handle.net/10419/191689
Authors: 
Raheem, Maruf A.
Ezepue, Patrick O.
Year of Publication: 
2016
Citation: 
[Journal:] CBN Journal of Applied Statistics [ISSN:] 2476-8472 [Volume:] 07 [Year:] 2016 [Issue:] 2 [Pages:] 77-99
Abstract: 
We present in this paper an alternative approach to determining and predicting the fluctuations in the daily prices and stock returns of a first-generation bank in the Nigerian Stock Market (NSM). The approach uses a three-state Markov to estimate the expected duration of the asset returns in states classified as rising (positive) (....), falling (negative) state (....) or stable (zero) state (....). Related goodness-of-fit tests show that the Markov model fits the data adequately with an error rate of approximately 0.1. The maximum expected lengths of successively being in either positive or negative regime is 4 days, while that of zero regime is 12 days, within any trading month of the study period (August 2005-Jnauary 2012). For the 2005-2009 period which encompasses post-2004 banking reform and the 2007-2009 global financial crisis, runs of zero returns dominate those of positive and negative returns about 59% of the time, indicating a lack of pronounced asymmetric effects in the bank's returns. The findings further reveal a minimum trading cycle of 7 days in February and a maximum cycle of 18 days in the months of May and October. The paper provides useful insights not only on the durations of returns in the three states, but on the Markovian transition probabilities among pairs of states which have implications for how investors could trade and invest in the bank stock or in a portfolio with bank stocks, if the same approach is used to characterise the returns dynamics of other banks in the NSM.
Subjects: 
Markov model
predictability
probability transition matrix
regime changes
stock returns
trading cycle
JEL: 
C01
C5
C12
C13
C58
Document Type: 
Article

Files in This Item:
File
Size
941.06 kB





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