Title: Testing structural breaks: A new self-normalization approach based on the adjusted sample range
Authors: Yongmiao Hong - Department of Economics at Cornell University (United States)
Brendan McCabe - University of Liverpool (United Kingdom)
Jiajing Sun - University of Chinese Academy of Sciences (China) [presenting]
Abstract: The aim is to test for structural breaks, and to propose a new self-normalization approach based on the adjusted range of the partial sum process. First, we extend the Kolmogorov-Smirnov test statistic using the adjusted range based self-normalization to test for a change in the mean. Second, we extend previous and introduce the G statistic, based on the adjusted range of the partial sum, which can accommodate multiple structural changes and structural changes in a multi-dimensional setting. Third, we extend the range-normalized KS and G statistics to consider structural changes in a general setting. Fourth, we extend the range-normalized KS and G test statistics to testing parameter constancy under a conditional autoregressive (CAR) framework. The focus on CAR follows from the superiority of the range-based self-normalization under conditions of persistent autocorrelation. However, parameter constancy tests developed can be easily generalized to other tests of parameter constancy under suitable conditions. We explore the statistical properties of these test statistics and conduct simulation and empirical studies. Our results show that the range-based test statistics are amongst the most reliable offering monotonic power when the nave self-normalization and kernel based long-run variance estimators fail to do so. The empirical studies also reveal that range-based test statistics are of great help in testing for structural breaks.