The Purchasing Power Parity puzzle: a sudden nonlinear perspective

Abstract

The aim of this study is to construct a simple nonlinear model for the US dollar–euro real exchange rate. The nonlinear model considered allows the adjustment towards long-run equilibrium to be sudden as well as smooth. It was found that the adjustment is sudden.

Notes

¹ Several papers have added terms in exchange rates to otherwise standard Taylor rules. See, for example, Clarida et al. (1998), Engel and West (2004).

² Cheung and Chinn (2001) found that at the six month horizon 81% of traders view PPP as irrelevant. At the long horizon only 40% of traders agree that PPP has some influence.

³ Obstfeld and Taylor (1997) use data measured relatively to the US after 1980.

⁴ Michael et al. (1997) use long time series spanning 1791–1992 and also post-war time series for bilateral US real exchange rates. Baum et al. (2001) examine bilateral US dollar CPI and WPI proxies over the post-Bretton Wood period. The data set in Taylor et al. (2001) comprises several bilateral real exchange rates against US dollar during the post-Bretton Wood period.

⁵ From 1979 to 1982, the Federal Reserve (FED) targeted non-borrowed reserves. Since then, the FED has followed a monetary policy rule which practically amounts to an interest rate targeting policy.

⁶ The results did not change if a linear trend was included in the Dickey-Fuller regression.

⁷ According to Eklund (2003) the size of the test is distorted when the value of δ₁ is close to −1 or 1. The size of the ADF test is distorted only when δ₁ is close to 1.

⁸ Critical values are used for 250 observations.

⁹ An ordinary F-test is used, since, as found by Granger and Teräsvirta (1993), an F-approximation works much better with small sample size than LM test with the asymptotic χ² distribution.

¹⁰ The p-values for the whole sequence of test are given only if the general linearity test (F) lies below 0.05.

¹¹ See a similar finding in Bec et al. (2002). They use a three-regime LSTAR model with the symmetry restriction.

¹² Taylor et al. (2001) is an exception (d = 1. For an example of a large d value, see Baum et al., 2001).