On the more powerful unit root test

Guide(s)

Bhattacharyya, Malay

Department

Decision Sciences

Area

Decision Sciences

University

Indian Institute of Management Bangalore

Place

Bangalore

Publication Date

3-31-2019

Year Awarded

March 2019

Year Completed

March 2019

Year Registered

June 2013

Abstract

In this thesis, we propose a unit root test based on Johnson SU density when the AR-coefficient is very close to one (local-to-unity) in the true data generating process. Through extensive simulations, we find that the proposed test, in finite samples, is as powerful as Dickey-Fuller test for normal errors and is significantly more powerful than many existing tests for non-normal errors. The proposed test has been applied on the Nelson and Plosser data set and on the nominal monthly interest rate of India. The proposed test performs better than several other tests available in literature. A strategy has been developed for unit root test that helps practitioners to decide whether they should use time trend in the model or drift in the model. This strategy has shown pretty good small sample performance, especially for the case with asymmetric errors.

Pagination

90p.

Copyright

Indian Institute of Management Bangalore

Document Type

Dissertation

DAC Chairperson

Bhattacharyya, Malay

DAC Members

Murthy, Rajluxmi V; Venkatagiri, Shankar

Type of Degree

Ph.D.

Comments

`

Relation

DIS-IIMB-FPM-P19-11

This document is currently not available here.

Request Access

Share

COinS