Detailed description of X-12-ARIMA seasonal adjustment of Serbian industrial production series

In this paper we seasonally adjust, step by step, monthly series of Serbian industrial production for the period 1994–2012, using the X-12-ARIMA method. In the pre-adjustment phase, we corrected the series for the automatically detected (using RegARIMA) effects of outliers, such as the level shift for NATO bombing, and the temporary change for the political changes of 5 October 2000 etc, as well as the (user-specified) ramp effect for the beginning of the global financial crisis. Using RegARIMA we also estimated that one working day less, because of weekends or holidays, reduces industrial production by 1.5%. These effects were also excluded from the series at this stage, before entering the next, X-11 phase. The final seasonal component in the X-11 phase, shows declining seasonal fluctuations of the series over the last 10 years, that is a lower seasonal factor for winter-autumn months and higher for summer months. Besides adjusting for the seasonal factor, we also adjusted the series for the calendar effect (number of working days and holidays), thus providing a more realistic picture on the movements in industry. Seasonally adjusted series contains all the outliers, with temporary ones being included in the irregular component, and permanent ones in the trend. The revisions analysis showed January as the least “reliable” month, that is the month for which seasonally adjusted data are revised mostly when new data are added to the series.