Asymmetric adjustment towards long-run PPP: Some new evidence for Asian economies

Abstract

This paper investigates relative purchasing power parity for a sample of nine Asian economies during the post-Bretton Woods floating exchange rate era. While most existing studies of purchasing power parity employ linear tests of non-stationarity or non-cointegration, we employ a new cointegration test, recently advocated by Enders & Siklos and Enders & Dibooglu, that tests for an asymmetric adjustment towards parity with respect to positive and negative real exchange rate deviations from parity. In most cases, we find that long-run purchasing power parity is most likely to hold with respect to positive deviations only.

Keywords:

• PPP
• cointegration
• asymmetries

Acknowledgment

We are grateful for the very helpful comments made by the Editor and two anonymous referees. The usual disclaimer applies.

Notes

¹New Keynesian macroeconomics provides models of wage and price rigidity based on optimising agents that can be used to justify the assertion that nominal monetary shocks lead to an asymmetric response from real output. Confirmation of asymmetries has been provided by a range of studies that include De Long & Summers (1988), Cover (1992) and Karras (1996) on the US and European economies, which suggest that real output is relatively more sensitive to negative rather than positive monetary shocks. Rhee (1995) finds evidence that equivalent positive and negative monetary shocks have an inflation-varying asymmetric effect on Korean output.

²See Sarantis (1999) and references contained therein.

³However, it can be pointed out that the STAR methodology implies that the speed of adjustment towards long-run PPP is positively related to the size of the real exchange rate shock while the threshold cointegration model implies a constant speed of adjustment.

⁴See Crownover et al. (1996) for an elaboration on this point.

⁵Hansen (1997) argues that if μ is a near unit root process, setting the Heaviside indicator using the change in μ can perform better than the specification using pure TAR adjustment.

⁶Hansen (1997) points out that the small sample properties of the OLS estimates for ρ₁ and ρ₂ have inflated standard errors and the convergence properties of the OLS estimates can be poor. This has implications for standard methods of inference concerning the individual estimates of ρ₁ and ρ₂. This is particularly a problem when the true value of the threshold is unknown.

⁷For example, in the case of Singapore, the point estimates of the ρ _i s are both negative and the φ_μ statistic for the null hypothesis ρ₁ = ρ₂ = 0 of 9.359 exceeds the 5% critical value that we have computed. The conventional F-test for symmetric versus asymmetric adjustment gives the value of 31.425, which strongly rejects the null of symmetric adjustment at the 1% significance level. The speed of adjustment of the real exchange rate is 24.2% per month with respect to a positive deviation from PPP and 0.9% per month with respect to a negative deviation. However, the coefficient on ρ₂ is insignificantly different from zero.