Cointegration tests of purchasing power parity

In recent work Im, Lee, and Enders (2006) use stationary instrumental variables to test for cointegrating relationships. The advantage of their approach is that the t-statistics are asymptotically standard normal and the familiar critical values of the normal distribution may be used to assess significance. Thus, the test avoids the nuisance parameter problem in single equation regressions for cointegration. Using an updated version of the data set developed by Taylor (2002), the ILE test is compared to three single equation alternatives in testing for purchasing power parity: An error correction model, autoregressive distributed lag model, and the Engle-Granger two step procedure. The regressions with instruments provide evidence supportive of PPP for some countries but the empirical results differ across tests and the choice of instrument can affect the results.