ABSTRACT

Empirical testing of the arbitrage pricing theory using observed economic and financial variables can be divided into three main steps: identifying the observed variables; generating the innovation terms of observed variables; and estimating their prices of risk. The first and second steps were presented in the last chapter. The second step, which deals with generating the unexpected changes, was the subject of the analysis in Chapter 4, where we presented two techniques for generating innovations from the observed variables. The first technique is used by Clare and Thomas (1994) and Clare, Priestley and Thomas (1999) and is based on using residuals from restricted autoregressive models as innovation terms. The second technique is proposed by Chen, Roll and Ross (1986) and is applied by Costa, Gardini and Paruolo (1999) and is based on using the residuals from the vector autoregressive (VAR) model as innovation terms. Based on the general and statistical properties of the two types of residuals that are generated by the two techniques, in the last chapter we concluded that the residuals that are generated from restricted autoregressive models have an advantage over those that are generated by vector autoregressive models. The objective of this chapter is to determine the economic and financial variables that command a risk premium in determining excess returns of listed shares in our sample of five countries, which is the third step in testing the APT model using the non-linear least square method and the reduced rank regression method of estimation.