The present paper calculates systematic risk within the context of the Capital Asset Pricing Model in order to investigate the significance of financial leverage. It develops a multinomial model with two theoretically predicted targets in the unleveraged/leveraged process, namely the proxy levered beta and an error term. This model allows us to address two issues. First, we empirically test two contradictory models, Modigliani and Miller (1958, 1963) and Miles and Ezzell (1985), regarding the correct rate to discount tax shields. Second, we investigate whether utilising the basic idea behind unlevered betas helps overcome the information shortfalls in calculating the cost of capital for unlisted firms. We find that the assumptions made by Modigliani and Miller are statistically more robust than are those made by Miles and Ezzell. Our results suggest that the use of the proxy levered beta to solve the lack of market information problem for both non-traded firms and individual business units is not misleading.