Endogeneity in Corporate Finance Empirical Research (In Portuguese)
32 Pages Posted: 23 Apr 2010
Date Written: April 20, 2010
This paper offers solutions to mitigate the problems associated with the endogeneity issues of explanatory variables faced by corporate finance researchers who analyze observational data via regression methods. Such problems may bias the causality inferences between variables, a fundamental goal in virtually all empirical works on corporate finance.
We identify the main methodological challenges faced by empirical researchers, as well as the main regression methods available in order to reduce the possibility of drawing incorrect causality inferences. Then, we develop and implement an original simulation model that reproduces the main characteristics of the observational data employed in most corporate finance researches.
As the main result, the computational experiments evidence that the most usually employed regression methods in corporate finance are generally inadequate and can lead to substantial incorrect inferences. On the other hand, they point to estimators based on Generalized Method of Moments applied to panel data as effective and feasible ways to eliminate or at least mitigate the endogeneity problems usually found in this line of research. The general implication is that both theoretical and simulation analyses lead to a questioning of the results of previous works on corporate finance, particularly the ones who solely employ ordinary least squares and fixed or random effects as a way for drawing causality inferences.
Note: Downloadable document is in Portuguese.
Keywords: Endogeneity, simulation, corporate finance, GMM
JEL Classification: C23, C63, G30
Suggested Citation: Suggested Citation