Smooth breaks and nonlinear mean reversion in real interest parity: Evidence from East Asian countries

ABSTRACT

This study aims to explore the empirical validity of the real interest rate parity (RIP) hypothesis for East Asian countries using Japan as the base country. To this end, we employ the recently proposed unit root tests of Christopoulos and Leon-Ledesma that account for both multiple smooth structural breaks of unknown form and nonlinear mean reversion in the series. Our empirical results uncover overwhelming evidences in favor of the RIP hypothesis for the whole countries in our sample. More specifically, through a Fourier approximation, it is observed that all real interest rate differentials display a mean-reverting behavior around an infrequently smooth-breaking mean, with the breaks being in accordance with the financial reforms and economic crises witnessed by the countries. Moreover, the degree of mean reversion appears to vary nonlinearly with the size of real interest rate appreciations and depreciations.

KEYWORDS:

• Real interest rate parity
• financial integration
• nonlinearity
• smooth structural breaks
• East Asian countries

JEL CLASSIFICATIONS:

• C22
• E40
• F36
• F40
• G01

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. See Neely and Rapach (2008), and Sirichand et al. (2015) for an extensive review of the RIP literature.

2. See Kaminsky and Schmukler (2003) for a more detailed discussion of the country-specific liberalization policies.

3. See Kapetanios et al. (2003) for further details.

4. As pointed out by an anonymous referee, almost all rids illustrated in Figure 1 are subject to sharp structural breaks as well. Since the FADF and the KPSS-type stationarity tests are designed to work best when breaks are gradual, it would be nice to perform an additional stationarity analysis for rids through unit root tests allowing for sharp structural breaks. As such, we employed Zivot and Andrews (1992) and Lumsdaine and Papell (1997) unit root tests, which allow for one and two endogenous structural breaks in intercept and/or trend terms, respectively. The results, available upon request, indicate stationarity only for Indonesia, Philippines and Korea at the 5 percent significance level. Despite being designed to capture sharp structural breaks in the series, it is noteworthy that the tests of Zivot and Andrews (1992) and Lumsdaine and Papell (1997) fail to reject the null of nonstationarity of rids for Singapore and China where the existence of sharp breaks is more prominent compared to other cases which appeared to be nonstationary according to the tests of Zivot and Andrews (1992) and Lumsdaine and Papell (1997).

5. Throughout our analysis the augmentation order of all unit root tests is chosen according to AIC. However, as underlined by an anonymous referee, the AIC criterion tends to overestimate the lag order in large samples. To avoid such a potential overestimating problem and to check for the sensitivity of our results to alternative information criteria we repeated all unit root tests by choosing lags according to the Bayesian information criterion (BIC) of Schwarz, test results remained almost unchanged.

6. Results from the LM-type linearity test are not presented here but available upon request.

7. We also performed the unit root test of Kılıç (2011), which does not take the smooth breaks into account but allows only for nonlinear mean reversion by forming the test regression (11) for rids. According to the results, available upon request, the null of nonstationarity cannot be rejected for South Korea, Singapore, China and Malaysia. This finding highlights the importance of accounting for nonlinearity and smooth breaks simultaneously and suggests that allowing for nonlinearity only offers a partial solution to non-rejection of nonstationarity of rids.