Testing for nonlinear unit roots in the presence of a structural break with an application to the qualified PPP during the 1997 Asian financial crisis

• Published date: 01 July 2018
• DOI: http://doi.org/10.1002/ijfe.1613
• Author: Kristofer Månsson,Olivier Habimana,Pär Sjölander
• Date: 01 July 2018

RESEARCH ARTICLE

Testing for nonlinear unit roots in the presence of a

structural break with an application to the qualified PPP

during the 1997 Asian financial crisis

Olivier Habimana

1,2

| Kristofer Månsson

1

| Pär Sjölander

1

1

Department of Economics, Finance and

Statistics, Jönköping International

Business School, Jönköping University,

Jönköping, Sweden

2

Department of Applied Statistics, College

of Business and Economics, University of

Rwanda, Kigali, Rwanda

Correspondence

Kristofer Månsson, Department of

Economics, Finance and Statistics,

Jönköping International Business School,

Jönköping University, Jönköping, Sweden.

Email: kristofer.mansson@ju.se

JEL Classification: C12; C15; C22; F31

Abstract

This paper applies Monte Carlo simulations to evaluate the size and power

properties in the presence of a structural break, for the standard Augmented

Dickey‐Fuller (ADF) test versus nonlinear exponential smooth transition

autoregressive unit root tests. The break causes the tests to be undersized,

and the statistical power considerably decreases. Moreover, the effect is intensi-

fied in small samples and very much increased for more persistent nonlinear

series. As a remedy, we modify the standard ADF and exponential smooth tran-

sition autoregressive unit root tests in order to adjust for a structural break. This

improves both the power and the size considerably, even though the empirical

size still is lower than the nominal one. More persistent series are more affected

by structural breaks, and the new tests are most powerful under the existence of

a rather persistent nonlinear data generating process (which is an empirically

relevant and common type of data generating process). The proposed tests are

applied to investigate mean reversion in the real effective exchange rates of 5

East and Southeast Asian countries, taking into account the structural change

in exchange rate regime brought about by the 1997 Asian financial crisis. The

empirical findings corroborate our simulation results; the modified more pow-

erful tests are able to reject the unit root in all 5 countries, whereas the tests that

do not consider the structural break could only reject in one of these cases.

KEYWORDS

exchange rates,Monte Carlo simulations, nonlinearity, qualified PPP,structural break, unit root test

1|INTRODUCTION

Ever since the seminal work by Nelson and Plosser (1982),

stationarity testing has been an obvious concern in econo-

metric analysis in applied macroeconomics. However,

there is a gap in previous research regarding assessing

the performances of nonlinear unit root tests in this

research field. It is a stylized fact that, given a robust sta-

tistical size, the power of nonlinear tests is larger and

more adaptable to different data generating processes

(DGPs). Nevertheless, recently compelling evidence of

the presence of nonlinearities in economic time series

has led to the developments of various nonlinear unit root

tests. The challenge (apart from structural breaks) has

been to develop tests that are able to distinguish linear

nonstationary processes from stationary nonlinear ones.

In this context, Kapetanios, Shin, and Snell (2003) pro-

pose the exponential smooth transition autoregressive

(ESTAR KSS) unit root test and demonstrate that it per-

forms better in terms of size and power especially when

Received: 8 September 2016 Accepted: 13 December 2017

DOI: 10.1002/ijfe.1613

Int J Fin Econ. 2018;23:221–232. Copyright © 2018 John Wiley & Sons, Ltd.wileyonlinelibrary.com/journal/ijfe 221