We find that an important programming error was made by Hobijn and Franses (2000) in conducting multivariate stationarity tests. The empirical results in their paper are subject to errors. Other studies that have used the coded algorithm of Hobijn and Franses are likely to suffer the same problem.
