Title: Specification of dynamic panel data models: An empirical application to corporate capital structure
Authors: Jan Kiviet - University of Amsterdam (Netherlands) [presenting]
Julio Pindado - University of Salamanca (Spain)
Ignacio Requijo - University of Salamanca (Spain)
Abstract: In corporate finance capital structure is modeled simply by partial adjustment. We demonstrate that this is rejected against more complete forms of dynamic adjustment. Also, usually only the endogeneity problems associated with the joint occurrence of unobserved firm effects and the lagged-dependent variable as explanatory factor are taken into account, without paying attention to possible endogeneity due to genuine simultaneity of the remaining regressors. We develop a model specification search strategy in which the actual form of the dynamic adjustments and possible relevance of instantaneous and lagged feedbacks from the dependent variable, with respect to the explanatory variables, are assessed from the data. We do this by applying the generalized method of moments technique to an ordered set of sequentially augmented restrictions on the parameters of a general dynamic model specification derived from the main capital structure theories. In the same process, we exploit and verify a sequentially augmented set of orthogonality conditions. At the various stages of this search strategy, diagnostic tests are used to decide on either the termination or the direction of the sequential further articulation of the imposed coefficient restrictions and adopted orthogonality conditions. For US firms, we obtain results that shed new light on the actual pattern and speed of dynamic adjustments and on the relevance of lagged and instantaneous feedbacks via other variables.