REAL EXCHANGE RATE DYNAMICS UNDER THE CURRENT FLOAT: A RE-EXAMINATION* by MICHAEL BLEANEY and STEPHEN J. LEYBOURNE University of Nottingham Augmented Dickey–Fuller regressions on pooled (but not individual) real exchange rates for the post-1973 period consistently reject the unit root null, even after accounting for cross-sectional dependence. The inference that the series is stationary, however, is not necessarily correct, because these tests strongly over-reject the null in certain circumstances, particu- larly when the series has a stochastic unit root. We find that bilateral real exchange rates against the US dollar have a stochastic unit root. Out-of- sample prediction exercises for an autoregressive model confirm these findings. 1 I  In a recent survey article, Rogoff (1996, p. 647) stated that ‘at long last, a number of recent studies have weighed in with fairly persuasive evidence that real exchange rates (nominal exchange rates adjusted for differences in national price levels) tend toward purchasing power parity in the very long run’. This evidence comes from single-series augmented Dickey–Fuller (ADF) tests on 80 years or more of data (Cheung and Lai, 1992; Lothian and Taylor,1996) and from pooled ADF tests on US dollar exchange rates for the post-1973 period (Jorion and Sweeney, 1996; Oh, 1996; Wu, 1996; Papell, 1997). These findings appear to have decided the debate about whether real exchange rates are stationary (I(0)) or integrated (I(1)) in favour of station- arity. This is a comforting conclusion for economists. Is it warranted? Sig- nificant notes of scepticism have been injected by O’Connell (1998) and Engel (2000). Engel shows that ADF tests may still reject the null of a unit root in long historical series when there is a substantial non-stationary element in the data. This is not inconsistent with economic theory, however, which allows for the possibility that the equilibrium real exchange rate varies over time. O’Connell criticizes pooled ADF tests that fail to allow for cross- correlations between real exchange rate series. Using a pooled ADF test based on a generalized least squares (GLS) procedure which corrects the test’s size The Manchester School Vol 71 No. 2 March 2003 1463–6786 156–171 *Manuscript received 30.3.01; final version received 7.5.02. © Blackwell Publishing Ltd and The Victoria University of Manchester, 2003. Published by Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK, and 350 Main Street, Malden, MA 02148, USA. 156