In a recent examination of the integrated nature of inflation, Culver and Papell (Journal of Applied Econometrics, 1997) applied a range of unit root and stationarity tests to data from a panel of 13 OECD economies. The results obtained were mixed. While little evidence of stationarity was detected using univariate methods, rejection of the unit root hypothesis was observed under panel data unit root testing, although rejection was found to be sensitive to cross-sectional variation. In this note the results of Culver and Papell are reconsidered in light of conditional heteroskedasticity detected in the inflation rate series. Using a more appropriate univariate testing procedure combining local-to-unity detrending and joint maximum likelihood estimation of a unit root testing equation and GARCH process, strong evidence in favour of stationarity is detected in 11 of 13 economies examined. In contrast to the univariate findings of Culver and Papell, the results obtained herein using an alternative univariate procedure provide evidence in support of their I(0) inference drawn using panel methods.
