“Beta risk estimation of companies listed on the Ghana stock exchange
Both researchers and practitioners have used beta as estimation for risk. Beta is usually estimated by running a regression on the market model. However, this estimation procedure can result in a variety of beta estimates. Some of the issues that can affect beta estimates are the measurement of returns, the choice of market index used, the length of estimation period and the sampling interval. The issue of normality, autocorrelation and the effect of thin trading, seasonality and stability also affect beta estimations. This project investigates these estimation issues using thirty-five companies. The empirical results generally confirm the evidence by various researchers in the literature review some which are Pike and Neale, Roll, French and Fama, Chan and Lakonishak, Amihud and Christen Damodaran. However, the tests for the effect of thin trading and the effect of seasonality reject the null hypothesis (Ho: Monday = Tuesday = Wednesday = Thursday = Friday ) According to Bradfield, detecting the beta stability is clouded by the fact that only estimates of beta are observable and that changing estimates do not necessarily imply stable underlying betas. Bowie and Bradfield (1997) found that the tests of beta stability are difficult to interpret on their own. Gombola and Kahi advocate that an OLS estimate of beta requires an estimation period during which the relationship between stock return and market return is stable. Without this stability, they suggest that an alternative for forecasting a time-varying relationship such as the general Bayesian adjustment process will be required.
A thesis submitted to the Department of Finance, Kwame Nkrumah University of Science and Technology, Kumasi in partial fullfilment of the requirements for the degree of Master of Business Administration (MBA) banking and finance.