Submission History & Decisions
Detecting Structural Change with Heteroskedasticity
PUBLISHED AS:
Mumtaz Ahmed, Gulfam Haider & Asad Zaman (2016): Detecting structural change with heteroskedasticity, Communications in Statistics - Theory and Methods, DOI: 10.1080/03610926.2016.1235200
Central Listing for above paper on AZ Articles: Ahmed, Haider, Zaman (2016) Detecting Structural Change
Above paper evaluates the rolling MZ (Maasoumi-Zaman) test for structural change, which was developed in the following paper: -- a final pre-publication draft of this paper is also attached below:
Maasoumi, Esfandiar, Asad Zaman, and Mumtaz Ahmed. "Tests for structural change, aggregation, and homogeneity." Economic Modelling 27.6 (2010): 1382-1391.
Central Listing for Above Paper: Maasoumi, Zaman, Ahmed: (2010) Tests for structural change
Changing Point and Parameter Instability with Heteroskedastic Models
Mumtaz Ahmed[1], Gulfam Haider[2], Asad Zaman[3]
Abstract
The hypothesis of structural stability that the regression coefficients do not change over time is central to all applications of linear regression models. It is rather surprising that existing theory as well as practice focuses on testing for structural change under homoskedasticity – that is, regression coefficients may change, but the variances remain the same. Since structural change can, and often does, involve changes in variances, this is a puzzling gap in the literature. Our main focus in this paper is to utilize a newly developed test (MZ) by Maasoumi et al. (2010) that tests simultaneously for break in regression coefficients as well as in variance. Currently the sup F test is most widely used for structural change. This has certain optimality properties shown by Andrews(1993). However, this test assumes homoskedasticity. We introduce the sup MZ test which caters to unknown breakpoints, and also compare it to the sup F. Our Monte Carlo results show that sup MZ test incurs only a low cost in case of homoskedasticity while having hugely better performance in case of heteroskedasticty. The simulation results are further supported by providing a real world application. In real world data sets, we find that structural change often involves heteroskedasticity. In such cases, the sup F test can fail to detect structural breaks and give misleading results, while the sup MZ test works well. We conclude that the sup MZ test is superior to current methodology for detecting structural change.
Key Words: Regime Shift, Heteroskedasticity, sup F, sup MZ and Monte Carlo Simulations.
JEL Classification: C15, C22, C32.
[1] Corresponding author, Assistant Professor, Department of Management Sciences, COMSATS Institute of Information Technology, Islamabad, Pakistan.
Email: mumtaz.ahmed@comsats.edu.pk, mumtaz.mumtazahmed@gmail.com
[2] PhD Econometrics student International Islamic University Islamabad, Pakistan and Lecturer at School of Management FAST NUCES CFD Campus. Email:haider.gulfam@nu.edu.pk
[3] Professor of Economics, Vice Chancellor, Pakistan Institute of Development Economics, Islamabad, Pakistan. Email: asadzaman@alum.mit.edu.pk