International Journal of Forecasting 17 (2001) 203–230 www.elsevier.com / locate / ijforecast Structural breaks, ARIMA model and Finnish inflation forecasts * Juha Junttila Department of Economics, University of Oulu, P .O. Box 4600, FIN-90014 Oulu, Finland Abstract Via the use of the rolling regression technique and a specific procedure for analysing strong structural breaks in a univariate time series model, we forecast the rate of future inflation in Finland for the time period of unregulated financial markets since the beginning of 1987. The identified structural changes in the data generating process (DGP) of inflation are labelled with both economic events and changes in the main leading inflation indicators. The final intervention model yields, in some cases, better forecasts than the pure rolling regression technique without identification of the strong breaks. When comparing the obtained forecasts with certain noncontinuous time series based on inflation expectation surveys with respect to actual future inflation, we find that the comparable point forecasts from our rolling regressions perform better than the corresponding point expectation proxies from questionnaires. When compared with the performance of the forecasts by the Research Institute of the Finnish Economy, the recursive procedure also produces more accurate forecasts.  2001 International Institute of Forecasters. Published by Elsevier Science B.V. All rights reserved. Keywords: AR(I)MA models; Structural breaks; Time variation; Forecasting 1. Introduction ported to invalidate the modern econometric testing procedures of, for example, unit root The role and effect of structural changes in analysis. By utilising the same set of several US the economy, economic models and data macroeconomic time series, the group of papers generating processes (DGPs) of economic vari- starting from Nelson and Plosser (1982), via ables themselves have been the subject of vivid Perron (1989) to Zivot and Andrews (1992) and discussion and research in economics for a long Newbold, Nunes and Kuan (1994) gives an time. The most frequently examined economic example of the fragility of unit root analysis variables include the rate of inflation and the when the possibility of structural changes is exogenously assumed events in the economy recognised (see also on this subject Dufour and supposed to have an effect on the DGP of Ghysels, 1996). inflation. Structural breaks have also been re- This study uses an ARIMA model in the forms of recursive and rolling regressions with monthly data for the period 1/1978–9/1996 to *Tel.: 1358-8-553-2916; fax: 1358-8-553-2906. E-mail address: juha.junttila@oulu.fi (J. Junttila). examine the impact of structural breaks on the 0169-2070 / 01 / $ – see front matter  2001 International Institute of Forecasters. Published by Elsevier Science B.V. All rights reserved. PII: S0169-2070(00)00080-7