Some issues in estimation of Beta

Abstract

In modern finance portfolio decisions are based on the interplay of risk and newlinereturn characteristics of the portfolio. An investor is generally faced with a newlinewide array of financial assets with different combinations risk and returns newlineto choose from. The combination of assets held in a portfolio depends on newlinethe risk an investor is prepared to take for a given level of expected returns. newlineThis makes quantification and accurate measurement of risk crucial for any newlineinvestment decision. newline newlineOne widely used measure of risk of a portfolio is the systematic risk or newlinethe beta of the portfolio. Beta is a measure of nondiversfiable portion of newlinethe portfolio risk and this is the risk for which an investor seeks returns. newlineThe concept of beta comes from the Capital Asset Pricing Model which newlinepostulates a linear model in which the expected asset returns are scaled up newlineover the expected market returns by a constant factor- beta. Beta is normally newlineestimated as the slope coefficient in the linear regression of asset returns over newlinethe market returns and is assumed to be constant over the estimation interval. newlineThere is overwhelming empirical evidence and sound economic arguments newlinein favour of time variations in beta. Various studies have conducted tests newlinefor time{variation in beta. This thesis tries an improvement over one of the newlinemost common methodlogy-the Kalman Filter, for testing for time variation newlinein beta. The results show that in its standard form Kalman Filter may give newlinebiased results. The new methodology is applied on data from the Indian newlinestock market to test for time{variation in beta. newline newlineThis thesis also empirically tests the validity of one of the results from newlinethe option pricing theory which postulates that beta is a negative function newlineof the risk free interest rates. The results show that theoretical predictions newlineare largely validated. newlineBetas are unobservable variable and little is known about there distributional newlinecharacteristics. There have been new developments in the \realized newlinevolatilityquot literature, where high{frequency intra{day data is used