An extensile method on the arbitrage pricing theory based on downside risk (D-APT)

Abstract

Purpose – The purpose of this paper is to propose a new and improved version of arbitrage pricing theory (APT), namely, downside APT (D-APT) using the concepts of factors’ downside beta and semi-variance. Design/methodology/approach – This study includes 163 stocks traded on the Malaysian stock market and uses eight macroeconomic variables as the dependent and independent variables to investigate the relationship between the adjusted returns and the downside factors’ betas over the whole period 1990-2010, and sub-periods 1990-1998 and 1999-2010. It proposes a new version of the APT, namely, the D-APT to replace two deficient measures of factor's beta and variance with more efficient measures of factors’ downside betas and semi-variance to improve and dispel the APT deficiency. Findings – The paper finds that the pricing restrictions of the D-APT, in the context of an unrestricted linear factor model, cannot be rejected over the sample period. This means that all of the identified factors are able to price stock returns in the D-APT model. The robustness control model supports the results reported for the D-APT as well. In addition, all of the empirical tests provide support the D-APT as a new asset pricing model, especially during a crisis. Research limitations/implications – It may be worthwhile explaining the autocorrelation limitation between variables when applying the D-APT. Practical implications – The framework can be useful to investors, portfolio managers, and economists in predicting expected stock returns driven by macroeconomic and financial variables. Moreover, the results are important to corporate managers who undertake the cost of capital computations, fund managers who make investment decisions and, investors who assess the performance of managed funds. Originality/value – This paper is the first study to apply the concepts of semi-variance and downside beta in the conventional APT model to propose a new model, namely, the D-APT.