Macroeconomic identification of the priced APT factors on the Johannesburg Stock Exchange

Abstract

Employing prespecified macroeconomic variables as potential priced factors, the Arbitrage Pricing Theory (APT) may be modelled as a non-linear seemingly unrelated regression with across equation restrictions. This portrayal allows for the simultaneous estimation of factor sensitivities and the risk premium associated with each factor. The following macroeconomic variables were tested as potential factors: unexpected movements in (rand) gold returns. (dollar) returns on the Dow-Jones Industrial Index, the term structure of interest rates and inflation expectations together with the 'residual market factor' of Burmeister Wall. Using iterated non-linear seemingly unrelated regression (ITNLSUR) estimation techniques, it was found that all of the above variables except for gold price risk are priced, that is, are associated with statistically significant risk premia.