Does PPP Hold Globally? Serial Correlation is More than You Think

Abstract

Recent literature reports that in estimating the purchasing power parity, potential biases may significantly modify the half-life of means reverting of long-run real exchange rates. This paper considers three kinds of potential biases on which empirical researchers focused most. In addition, we accept Levin, Lin and Chu (2002) dynamic panel data estimation method which is alternative to the classical dynamic panel data estimating approach. Instead of using the same length of all regressors, this modification allows each time series to choose the number of lagged terms independently according to the time series data itself. In other words, we check the significance of regression coefficients to let each time series tell how many lagged terms it need to eliminate the serial-correlation which is common in most macro-economics data. We reexamine 77 world major countries and economies’ CPI based real exchange rates for post Bretton Wood period. With a bias adjusted half-life estimation of approximate 8 months, this paper illustrates an interesting phenomenon that serial-correlation could be an important resource of the PPP puzzle.