Multivariate Regression
It is not appropriate to include all of the alternative proxy variables in our It is not appropriate to include all of the alternative proxy variables in our regression. Not only would many degrees of freedom be used up, but the model would be highly collinear. The method used to select the variable proxy measures was as follows. A series of OLS regressions were carried out which included all of the explanatory variables, but included only one of the alternative proxies in any one single regression. This procedure identified those proxies with the highest explanatory power — BETA, CAPGEAR, SALES, OOCI and MOSQ — which were then used in all Subsequent analysis. These are the proxy measures listed in Table l. Details of the alternative proxy measures used are given in note 3.