Robust cross sectional dependence test in panel regression model

Md. Shariff, N.S. and Baharumshah, A.Z. (2009) Robust cross sectional dependence test in panel regression model. In: International Conference on Robust Statistics, 14-19 Jun 2009, Parma, Italy. (Submitted)

robust_cross_sectional.pdf - Submitted Version

Download (4MB)


The central focus in most recent studies on panel data is on the issue of cross sectional dependence; there exist correlations between different groups (cross section) of innovations in the panel. In the presence of cross sectional dependence, we can no longer use the general assumption of independence across units for disturbances. Thus, it is important to test for cross sectional dependence before modeling and estimating panel data takes place in order to avoid the misspecification of the model which subsequently results in invalid tests, bias, inconsistency and inefficiency in parameter estimates. Difficulty arises when (i) disturbances are correlated across cross section; (ii) data are subjected to outliersl shocks. It has been reported that the main possible causes are global shocks and unobserved factors that affect the cross sectional dependence (see Cerrato (200)). However, in reality, it is also possible to have local shocks which affect innovations in the model. The question that usually arises is whether cross dependencies does exist among innovations in the presence of outliers. The usual tests based on Breusch and Pagan (1980) and Pesaran (2004) uses residuals obtained from Ordinary Least Square Fit (OLS) which are subject to influence of outliers. This paper propose robust cross dependency tests of Breusch and Pagan (1980) and Pesaran (2004) to aid cross sectional dependence test in the presence of outliers. Here, robust regression (RREG) model is employed to capture the outliers' effects and filter outlying observation by down weighting the spurious data. In this paper, we compare the performances of the Breusch and Pagan (1980), Pesaran (2004) tests as well as the proposed methods under varying degree of cross section dependencies via simulation studies. Our resJb have shown that robust approaches are able to produce reliable test of cross sectional dependency under the conditions considered.

Item Type: Conference or Workshop Item (Paper)
Uncontrolled Keywords: Panel Regression, Cross Section Dependence, Local Shocks, Robust Regression, Huber function, Diagnostic Tool.
Subjects: Q Science > QA Mathematics
Divisions: Faculty of Science > Institute of Mathematical Sciences
Depositing User: Mr. Mohd Samsul Ismail
Date Deposited: 21 Jul 2014 00:46
Last Modified: 21 Jul 2014 00:46

Actions (login required)

View Item View Item