HETEROGENEOUS PANEL UNIT ROOT TESTS AND PURCHASING POWER PARITY by KUL B. LUINTEL Brunel University Doubts have been raised recently on the ¢ndings of panel studies of purchasing power parity (PPP) on the grounds that they ignore serial correlation and cross-sectional dependence, and consequently su¡er from severe size biases and loss of power. We implement an alternative panel unit root test that controls for serial correlation and cross- sectional dependence as well as the heterogeneity of dynamics and error variances across groups, and ¢nd strong support for PPP. Our ¢ndings are consistent with other recent studies on the subject. " Introduction Purchasing power parity (PPP) is one of the most widely tested propositions in economics, not least because it remains one of the key building blocks of theoretical models in open economy macroeconomics. However, the empirical evidence is far from conclusive. Studies based on microeconomic data wholly reject PPP. Time-series tests on recent £oat data also to a large extent reject PPP. 1 Rogo¡ (1996, p. 655) calls the former set of evidence an `abject failure' of the law of one price and the latter an `embarrassment' to the proponents of PPP. 2 It is hypothesized that the lack of empirical support for PPP during the recent £oat is due to the low power of the tests, ascribed to the short data span. 3 In response, one strand of literature examines PPP using long- span data and time-series methods and the other resorts to panel tests. Both approaches provide enough data points for powerful tests and produce results favourable to the PPP hypothesis. Long-span time-series studies that analyse 60^700 years of data ¢nd, in most cases, slowly mean reverting real exchange rates, which implies that PPP holds in the long run. 4 Likewise, cross-country panel studies also ¢nd mean reverting real ß Blackwell Publishers Ltd and The Victoria University of Manchester, 2001. Published by Blackwell Publishers Ltd, 108 Cowley Road, Oxford OX4 1JF, UK, and 350 Main Street, Malden, MA 02148, USA. 42 The Manchester School Supplement 2001 1463^6786 42^56 1 For further details see, among others, Taylor (1995), MacDonald (1995) and Rogo¡ (1996) who provide excellent surveys of the empirical literature on PPP. 2 The theoretical possibility of a random walk real exchange rate is shown by Adler and Lehmann (1983), among others. However, for a criticism of this view see Diebold et al. (1991). 3 On this point see Frankel (1990), Frankel and Rose (1996) and Lothian and Taylor (1996). 4 Abuaf and Jorion (1990), Diebold et al. (1991), Grilli and Kaminsky (1991) and Lothian and Taylor (1996) are some of the long-horizon studies which support PPP. On the other hand, Froot and Rogo¡ (1995) do not ¢nd mean reverting real exchange rates even with long-span data.