Parity reversion in the Asian real exchange rates: new evidence from the local-persistent model

Abstract

This article investigates the time-series properties of 13 Asian real exchange rates (RERs) vis-à-vis the US dollar. The half-life point estimates drawn from the local-persistent model are all less than 2 years, with a finite upper bound. There is no evidence to indicate that the Asian financial crisis has altered the speed of the purchasing power parity (PPP) adjustments. We find that the persistence of RERs over the last three decades remains unchanged in majority of the cases. Given the fairly rapid speed of adjustments and their corresponding confidence intervals, we conclude that the PPP puzzle does not exist in these countries.

Keywords:

• Asian countries
• half-lives
• local persistent process
• real exchange rate
• crisis

JEL Classification:

• E30
• F31
• C23

Acknowledgment

We would like to thank the Editor of this journal for useful suggestions.

Notes

¹ As forcefully argued by Lothian and Taylor (1997), detection of unit root in RER may be due to issues associated with the ADF test that can have low power in small samples. To address the small sample problem, researchers use long horizon data or panel data to provide stronger rejections of the unit-root hypothesis.

² Long-run PPP means more than the absence of a unit root. It means a sufficient degree of mean reversion in PPP at the horizon of interest. For example, Kilian and Zha (2002) emphasize that test of unit-root null hypothesis may be unable to provide guidance as to whether the PPP assumption should be abandoned in economic modelling or not.

³ If parity holds continuously, then the volatility of the nominal exchange rate should not be greater than the volatility of the relative national prices, that is, 1–2 years.

⁴ For more on the stylized facts on the behaviour of the RER in developed and developing countries, see Drine and Rault (2008).

⁵ Near unit-root has long been recognized in the PPP literature. Based on the fractional integration framework, Cheung and Lai (2001) reject an exact unit root in favour of a near unit-root for the yen-based RERs. The process has an appealing feature to describe both large swings and mean reverting dynamics simultaneously. The implication of this finding (near unit-root process) is that a long data span, much more than the one used in the current study, is necessary to validate PPP using conventional methods.

⁶ The United States is usually chosen as the reference country mainly because of its importance in terms of international trade and investment in Asia.

⁷ Several papers (e.g. Sekioua, 2008; Kim and Lima, 2010) have highlighted that an evidence rejection of the unit-root null hypothesis could still be inconsistent with a stationary of RER, which has highly persistent shocks.

⁸ Evidence of near unit-root in various RERs has been frequently reported in the literature; see the paper in Kim and Lima (2010). The significance of such findings is that the conventional bootstrapped band is invalid when the data are generated by a near unit-root process.

⁹ Panel tests that do not allow for the fact that cross-sectional dependence may yield seriously biased estimates, especially if the degree of cross-sectional dependence is large. For more detail discussions on the presence of cross-sectional correlation across the Asian currencies (due to cross-country dependence); see Matsuki and Sugimoto (2013).

¹⁰ Ozdemir and Cakan (2010) highlight the persistency of the yen-based rates differs from the US dollar-based rates, with the latter having much more rapid speed of adjustment to PPP.

¹¹ Hegwood and Nath (2014) find that linear break models are much superior in detecting mean reversion. With only two breaks, they obtained unbiased half-life estimate of about a year (compared to 8 years without break) for the bilateral RERs between India and 16 of its trading partners.

¹² To our knowledge, the study by Astorga (2012) is the only one that examines the half-lives of developing countries using long horizon data (1900–2005). Besides that, the author also considers the role of fundamentals driving RER equilibrium. There are problems with longer span data that cut across the Bretton Woods era and flexible exchange period. Although PPP may hold for the entire sample period, the speed of convergence can be very different under different exchange regimes.

¹³ It has been documented that OLS autogressive (AR) estimators can be severely biased when the stochastic process is highly persistent, regardless of sample size. As discussed in Lima and Xiao (2007), the local persistent process is in between the stationary and unit-root extremes, similar to fractionally integrated process. They also show that if RER is modelled as a fractionally integrated process, economic shocks are not totally absorbed even after 30 years, compared to about only 3 years if the series is modelled as a local persistent process.

¹⁴ The data for TWN and CHN are drawn from Datastream.

¹⁵ Outliers could influence the outcome of standard unit-root tests. Several studies have shown that the over-rejection of the unit root is largely due to the presence of influential observations.

¹⁶ Factors that can account for cross-sectional differences in RER persistence include geographic proximity, market size, inflation and government spending. Previous work confirmed that there is an inverse relationship between income level and shock persistence (e.g. Cheung and Lai, 2000).

¹⁷ It is widely acknowledged that imposing parameter homogeneity in a panel setting will potentially bias the estimates of the slope coefficient, which may affect the true speed of the mean reversion (positive or negative). Imbs et al. (2005) refer to this as ‘aggregation bias’.

¹⁸ The degree of cross-dependence as well as the heterogeneity in panels of exchange rates affects the outcome of unit-root test and the speed of mean reversion. Homogeneous estimation of the mean reversion parameters exhibits potentially large biases.

¹⁹ The constructed CIs are much narrower compared to Baharumshah et al. (2008, 6 Asian), who used the ADF approximation, which did not correct for the downward trend of the conventional estimators.

²⁰ Unreported results show that RER in all Asian countries follows a stationary process in both the pre- and post-crisis periods. It is tempting to conclude that there is no clear relationship between country characteristics and the validity of PPP. It should be mentioned that the sample countries (N) might be too small to reach such a conclusion.

²¹ We generate the Hansen (1999) grid bootstrap CIs using the R programs, which are available from Hansen’s homepage at http://www.ssc.wisc.edu/~bhansen/progs/restat_99.html.

Additional information

Funding

Financial support from Universiti Putra Malaysia [grant number GP-IPB/2014/9440900] is acknowledged.