Abstract
In contrast to the capital structure propositions in Modigliani and Miller (1958, 1963), most empirical research has not been able to find a positive association between stock returns and financial leverage. Several reasons have been proposed for this, including suggestions about omitted risk factors or market mispricing of financial risk. However, the observations can also be due to the statistical models not properly representing the relation between stock returns and the return risk associated with leverage. In line with the well-known leverage formula, a statistical model should allow for a twofold return effect of leverage – a positive amplifying operating risk effect (corresponding to the basic idea of financial risk) and a negative interest cost effect. Modelling this twofold effect explicitly, we perform a number of tests of the association between stock returns and leverage with US data from 1966-2017. With the enterprise book-to-market as an indicator of operating risk, we find a robust positive regression coefficient of our variable of financial risk (depicting the amplifying risk effect of leverage) and a negative coefficient of the leverage variable. Including additional measures of operating risk and controls for negative leverage, we find that the negative leverage coefficient in the main is due to the credit risk premium of financial debt, and that this effect is significant only for commercially non-viable firms. However, in addition to the modelling in Modigliani and Miller (1958, 1963), our tests indicate that there might also be an additional, statistically driven, effect of leverage reinforcing the negative regression coefficient of leverage.