Median Regression for SUR Models with the Same Explanatory Variables in Each Equation

• Activity Number: 584
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Zangin A. A. Zeebari*+ and Ghazi Shukur
• Companies: Jonkoping International Business School and Jonkoping International Business School
• Keywords: Median SURE regression ; robustness ; efficiency ; SURE models

Abstract

In this paper we introduce an interesting feature of the Generalized Least Absolute Deviation (GLAD) method, that contrary to the collapse of GLS estimation of Seemingly Unrelated Regression Equations (SURE) to the OLS estimation when the same regressors are common between equations, the estimation of the proposed methodology does not collapse to the equation-by-equation LAD estimations. This is important since contrary to the least squares method, one can take advantage of efficiency gain due to cross-equation correlations even if the system includes the same regressors in each equation. Given the existence of systemwise medians of the errors from different equations when the cross-equation errors are correlated and regressors are the same for each equation, we show by a Monte Carlo experiment, an empirical study and analytical results that the proposed method does not collapse to the LAD estimation and it is more robust and more efficient than LAD estimation of the separate median regression equations and GLS method as well.