Decisions in North America – Panel Data Analysis and Endogeniuty Concern

[pic 1]

[pic 2][pic 3][pic 4]

Introduction

        The goal of this paper is to add insight into the factors that determine a firm’s optimal capital structure. This model extends current research by using an instrumental variable and 2SLS to overcome the endogeneity problem. Furthermore, is uses panel data which provides a deeper analysis into capital structure determinants when compared to a cross sectional or a time series regression. The data set is also extremely recent, 2005 – 2015, and adds interest coverage ratio which adds further insights into the empirical determinates of capital structure. Ozkan’s (2001) Gaud’s (2005) used firms from the (United Kingdom) UK and Switzerland use a dynamic panel data set and correct for endogeneity but this paper focuses on North American companies only which could have different results. Finally, only firms with assets over $50,000,000 are used as increases the effectiveness of the instrumental variable which will be discussed further in later sections.

Literature Review

        The Modigliani and Miller (1958) paper shows that a firm’s capital structure decision does not have an impact on the firm’s value, subject to constraints such as the perfect market assumptions. Since then, capital structure has been extensively studied in academia. Most of the emphasis has been putting on relaxing the Modigliani and Miller (1958) assumptions. For example: Jensen and Mickling (1976) take in account agency costs, Stiglitz (1972) and Titman (1984) study bankruptcy costs, Modigliani and Miller (1963) analyze the effects of personal taxes and Myers (1984) looks at information asymmetries.

        The two predominant theories of corporate capital structure are the “tradeoff theory” hypothesis and the “pecking order theory”. The tradeoff theory, developed by Myers and Majluf (1984), states that firms will optimize their choice of capital structure when the marginal benefit of debt equates the marginal cost. Marginal benefit of debt can be calculated using the tax shield, disciplinary roles of debt, less dilution of ownership when compared to equity financing, debt also suffers less form informational costs when compared to equity. Marginal cost is comprised of bankruptcy costs, lower credit ratings, increased cost of equity stemming from a higher required rate of return and agency costs between mangers and shareholders. According to this theory, managers make their capital structure decisions in order to minimize the costs mentions above. The tradeoff suggests the firm will prefer internal financing as opposed to external financing and debt as opposed to equity.  The pecking order theory also ranks the financing options in the same order.

        Rajan & Zingales’ empirical analysis (1995) investigates the determinants of a firm’s capital structure choice by analyzing the financing decisions of public firms in major industrialized countries. The model they propose for investigating this relationship utilizes a cross-sectional OLS regression of tangible assets, return on assets, market value to book value ratio and log sales on firm leverage.  However, this model does suffer from endogeneity and the use of a cross sectional regression as opposed to panel data limits the depth of analysis into the determinants of capital structure. Rajan & Zingales’ model, which uses OLS, will be biased and suffer from endogeneity if the error term and explanatory variables are correlated (Cov(xi,ei) ≠0). Given that researchers are more interested in causal analysis than just correlation analysis, it is clear to see why endogeneity presents a real concern for practitioners and academics. The usual suspects that cause endogeneity are in the form of model misspecification via omitted variable(s), measurement error and simultaneity. Though endogeneity remains an ongoing concern for any researcher, if a model is inappropriately specified, we know that with a clever identification technique OLS regressions can be unbiased. For instance, Bennedsen, M., Neilsen et all (2007) overcame their model’s endogeneity problem by incorporating an instrumental variable into their OLS regression. In addition to the above authors, Acemoglu, Johnson and Robinson (2001) were also able to identify an instrumental variable for their paper, “The Colonial Origins of Comparative Development”. Conversely, if an IV cannot be found, researchers may also turn to other techniques such as proxy variables in order to uncover causal relationships.