Forecasting the beta coefficient in a market characterised by thin trading

Abstract

The major goals of this work, which are the following: 1) We are interested in testing if the observed in many markets non-stationarity of betas through time, is still observable in a small market characterized by thin trading, the Athens Stock Exchange (ASE). For this purpose we are going to use the two methods originally proposed by Chawla (2001) of introducing and additional time variable in the classical OLS model and of using dummy variables to measure the change of the slope over time. 2) After establishing the “non-stationarity” tendency, we are interested in testing the forecasting ability of Blume’s and Vasicek’s methods to predict betas along time and conclude on which is the most appropriate amongst the two of them. 3) Moreover, we are going to evaluate the significance of a “correcting for thin-trading” technique in the estimation of beta and whether there is an improvement or not in the forecasting ability of the Blume’s and Vasicek’s methods to predict betas along time, when using as historical data, betas estimated by the a correcting procedure instead of estimating them by the classical OLS method. For this purpose we have chosen the Scholes’ & Williams’ methodology. 4) Finally, we are interested in checking whether the frequency of stock data collection has a significant impact (i.e. interval effect) on the forecasting ability of stock betas, by comparing the results achieved with daily and monthly data.