Title: MOSUM based test for variance change in panel data
Authors: Man Wang - Donghua University (China) [presenting]
Abstract: The problem of testing for common variance changes in panel data model is considered. We propose a test based on moving sum (MOSUM) of squares. Asymptotic distribution of the test statistic is derived under the null hypothesis of no change and the consistency of the test is also established. We conduct several simulation studies, which show that the proposed test performs better than some existing CUSUM based tests when the magnitude of variance change is small. Additionally, in the presence of multiple changes, the variances of the panels may increase at one change point and decrease at another. The increase and decrease in variances may cancel each other out and result in loss of power of the CUSUM based test method. Under this circumstance, the new MOSUM based test has obvious superiority over these CUSUM based test, as shown in the simulation.