Abstract
AbstractThis paper tests the PPP hypothesis for the South African rand/US dollar real exchange rate using a fractional integration framework. The results suggest that the real exchange rate of the South African rand with respect to the US dollar is a highly dependent variable with an order of integration very close to 1. This finding is not affected by the data frequency considered (daily, weekly or monthly). Also, there appears to be a single break in December 2001 (possibly corresponding to a change in the monetary policy framework), with the unit root null being rejected in favour of d > 1 for the periods before the break, but not afterwards. Thus, our results strongly reject the PPP hypothesis for the South African rand/US dollar rate across data frequencies, since shocks are found to affect the exchange rate forever.
Highlights
Purchasing Power Parity (PPP) is a central tenet in international economics
Others have applied unit root tests to real exchange rate. Such tests have been found to be unable to distinguish between random-walk behaviour and very slow mean-reversion to the long-run equilibrium level, as in small samples they have very low power against alternatives such as trend-stationary models (DeJong et al, 1992), structural breaks (Campbell and Perron, 1991), regime-switching (Nelson et al, 2001), or fractionally integration (Diebold and Rudebusch, 1991; Hassler and Wolters, 1994; Lee and Schmidt, 1996)
At times they exhibit erratic behaviour, suggesting the presence of endemic instability; adjusting the residuals for non-normality and heteroscedasticity using a wild bootstrap method attenuates this type of behaviour considerably, and the latter disappears almost completely if panel tests are performed, the evidence for PPP becoming much stronger
Summary
Purchasing Power Parity (PPP) is a central tenet in international economics. It is assumed to hold continuously in flexible-price models of the exchange rate, whilst in sticky-price ones it is a long-run property, temporary deviations from the long-run equilibrium being possible. The results are displayed in the first row in when including an intercept or an intercept with a linear time trend, the estimated value of d is around 0.978 and the unit root null is rejected in favour of slow mean reversion.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
