Multicollinearity in applied economics research and the Bayesian linear regression

Authors

  • Eric EISENSTAT Department of Economics, University of California, Irvine, United States

Keywords:

multiple linear regressions, classical normal regression, collinearity, multicollinearity, classical inference, subjective probability, Bayesian linear regression, prior information, posterior distributions, simulation

Abstract

This article revises the popular issue of collinearity amongst explanatory variables in the context of a multiple linear regression analysis, particularly in empirical studies within social science related fields. Some important interpretations and explanations are highlighted from the econometrics literature with respect to the effects of multicollinearity on statistical inference, as well as the general shortcomings of the once fervent search for methods intended to detect and mitigate these effects. Consequently, it is argued and demonstrated through simulation how these views may be resolved against an alternative methodology by integrating a researcher’s subjective information in a formal and systematic way through a Bayesian approach.

Downloads

Published

2016-04-13

How to Cite

EISENSTAT, E. (2016). Multicollinearity in applied economics research and the Bayesian linear regression. Annals of Spiru Haret University. Economic Series, 9(1), 47–58. Retrieved from https://anale.spiruharet.ro/economics/article/view/914

Issue

Section

ACADEMIA PAPERS