Content area
Full Text
1. INTRODUCTION
The evidence presented in finance literature strongly suggests that the costs associated with the adoption of a non-optimal capital structure might be substantial. A non-optimal capital structure decision, for example, might result in higher direct and indirect economic costs to firms in form of lower stock price (Masulis (1983), DeAngelo and Masulis (1980)), higher cost of capital and lost growth opportunities (Myers and Majluf (1984)), increased probability of bankruptcy (Warner (1977,1983), (Castanias (1983)), higher agency costs (Jensen and Meckling (1976), (Barnea, Haugen and Sunbet (1985)), and possible wealth transfers from one group of investors to another. The underpinnings of the modern capital structure theories are robust and intellectually appealing, and consequently, this subject is able to maintain a sustained and an abiding interest of researchers even after nearly five decades of intense research in this area.
The capital structure theories forwarded in the literature are predicated upon the existence of imperfections in the capital market such as taxes, bankruptcy and agency costs, asymmetric information between insiders and outsiders and signaling costs.1 Modigliani and Miller (1958) proved that in a perfect capital market, there is no relationship between the quantum of debt held by a firm and its value and moreover, the debt financing decisions have no bearing on the values of firms (and cost of capital). However, the existence of optional debt ratios for firms have been contended by numerous authors. Miller (1977) has argued in favor of the existence of optimal debt for the economy as a whole if not at a micro level. But interestingly, neither the investors nor managers can observe directly the optimal level of debt for a firm nor can they convincingly claim that the observed debt ratio is an optimal one. Empirically, it is infeasible to test either the existence of optimal capital structure for firms or the effect of debt on the value of the firm.
The main purpose of this paper is to revisit the capital structure decisions using the neural networks methodology. More specifically, the study compares the results derived from the classical linear statistical methods with neural networks results in explaining the observed cross-sectional debt ratios. Additionally this paper attempts to determine if the neural networks in general, have promising applications to capital...