In this paper we focus on the post Bretton Woods period and analyze whether a PPP relationship holds and what is the speed of adjustment to it. We adopt a multivariate system approach in which, initially, we test for cointegration and then we try to identify a cointegration space in which we have the PPP relationship (the "Johansen approach"). The studies that have adopted this approach have always rejected the PPP in favour of a long run relationship between the real exchange rate and the interest rate differential. On the contrary, our conclusions are in favour of the PPP for all the cases considered when we allow for a structural break in the data. We arrive to this conclusion, after having identified the cointegration space in two different ways: one in which we have the PPP as a cointegrated vector and one in which the real exchange rate plus the interest rate differential is a cointegrated vector. Adopting a dominance criterion we choose the former identification. We al so address the Rogoff's (1996) puzzle on the excess volatility of real exchange rate and the slow convergence to PPP. On the basis of persistence profiles obtained from a constrained VECM, we don't find any evidence in favour of the puzzle since we estimate a relatively fast speed of adjustment.