A new analysis of the determinants of the real dollar-sterling exchange rate: 1871-1994

Nonlinear models of deviations from PPP have recently provided an important, theoretically well motivated, contribution to the PPP puzzle. In recent work the equilibrium level has been modelled either as constant or as time varying with very similar statistical fits and very different economic implications. The high persistence of both PPP deviations and the proxy variables for the equilibrium real rate might create a problem of spurious coefficient significance. This paper investigates the possibility of spurious regression within nonlinear models of PPP. Monte Carlo experiments show that standard critical values are not appropriate in such a context. To illustrate we consider the real Dollar-Sterling exchange rate over the period 1871-1994. Due to many exchange rate regime changes over the sample period we employ a Bootstrap methodology that preserves the original structure of the estimated residuals and obtain new critical values of the coefficient estimates. A nonlinear (ESTAR) process with a time varying equilibrium proxied by relative wealth and relative income per capita seems to parsimoniously fit the data. Our results provide further evidence for the nonlinear model with a shifting equilibrium and the implied speed of adjustment is found to be substantially faster than previously reported in the literature.