Copious Structural Shifts in Exchange Rates of the South African Rand (Post-1994)

Abstract

There is a theoretical case for real exchange rates to be stationary, but conventional unit root tests generally find nonstationarity in most economic data expressed in nominal terms; exchange rates in particular. Perron (1989) questioned the latter interpretation on the basis that the presence of a unit root may be a manifestation of not allowing for structural change - a finding reaffirmed later by Zivot and Andrews (1992) and Clemente, Montanes, and Reyes (1998) when single and double sudden and gradual endogenous breakpoints are accounted for in unit root tests. This paper considers testing for structural breaks and unit roots - in the presence of structural shifts - in the univariate data generating process (DGP) of the key nominal foreign exchange rates of the South African rand. Additionally, the connexions between the timing of the structural shifts and important economic and noneconomic events are explored. The key findings show overwhelming support for structural shifts in the DGP and nonstationarity across all exchange rates examined - even after accounting for structural change. The convergence of the t-statistics for the yen/rand towards their critical values when at most two breaks are allowed for in the unit root tests is also a significant discovery; suggesting that stationarity is a possibility contingent on the sample period selected and the true number of breaks incorporated in the tests.