Time series adjustment in Austria

Transcription

1 We are moving Time series adjustment in Austria information Workshop II, 2 5 December 28, Vienna Statistics Austria S T A T I S T I C S A U S T R I A 1 Overview Background Basic idea Method used (X12-Arima) Adjustment of Austrian data on nights spent Problems & conclusion S T A T I S T I C S A U S T R I A 1

2 Background - seasonality in Austria - nights spent Development of nights spent in Austria in 27 18,, 16,, 14,, 12,, 1,, 8,, 6,, 4,, 2,, - Feb-7 Mar-7 Apr-7 May-7 Jun-7 Jul-7 Aug-7 Sep-7 Oct-7 Nov-7 Dec-7 Total Residents Non-residents S T A T I S T I C S A U S T R I A 3 Background - seasonality in Austria - employment Employment in Austria in all tourism relevant industries 1 ) in Share in % Feb-7 Mar-7 Apr-7 May-7 Jun-7 Jul-7 Aug-7 Sep-7 Oct-7 Nov-7 Dec-7 Total Full-time Part-time 1) ÖNACE 55, 6, 61, 62, 63 and 92 S T A T I S T I C S A U S T R I A 4 2

3 Background - seasonality in Austria - employment Employment in Austria in the accommodation and food service activities in Share in % Feb-7 Mar-7 Apr-7 May-7 Jun-7 Jul-7 Aug-7 Sep-7 Oct-7 Nov-7 Dec-7 Total Accommodation Food service activities S T A T I S T I C S A U S T R I A 5 Background - seasonality in Europe Nights spent: Ratio peak month in relation to lowest month in hr Croatia gr Greece bg Bulgaria cy Cyprus is Iceland Source: Eurostat, Tourism Statistics it Italy ro Romania mt Malta at Austria pt Portugal ie Ireland es Spain hu Hungary no Norway dk Denmark eu27 European Union S T A T I S T I C S A U S T R I A 6 3

4 Basic idea Main problem: Month-to-month comparisons not meaningful due to seasonal fluctuations Comparisons with the same month of the previous year not meaningful because intervening 11 months not taken into consideration Solution: Seasonal and calendar adjustment Uncovering of new developments Reliable interpretation of data possible!!! S T A T I S T I C S A U S T R I A 7 Basic idea decomposition model Observations can, due to varying, not overlapping causes, be broken down into a number of independent, not directly observable components! Trend long-term direction Economic cycle cyclical components with periods longer than one year Season seasonal fluctuations during a year Calender calendar irregularities Irregular component random fluctuations S T A T I S T I C S A U S T R I A 8 4

5 Basic idea Seasonal effect Different months vary: e.g. Summer vs winter, January 31 days, February 28 days, Calendar effect Same months vary: e.g. Number of working days in January, leap year effect, shifting of holidays (easter effect), S T A T I S T I C S A U S T R I A 9 Method X12-ARIMA Seasonal adjustment software Features: Extensive time series modeling and model selection capabilities for linear regression models with ARIMA errors (RegARIMA models) Wide variety of seasonal and trend filter options Diagnostics of the quality and stability of the adjustments achieved under the options selected S T A T I S T I C S A U S T R I A 1 5

6 Method X12-ARIMA Nights spent: data 25,, 2,, 15,, 1,, 5,, Jan-9 Jan-91 Jan-92 Jan-93 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan- Jan-1 Jan-2 Jan-3 Jan-4 Jan-5 Jan-6 Jan-8 Comparisons: with the previous year calendar adjustment with the previous month seasonal and calendar adjustment S T A T I S T I C S A U S T R I A 11 Method X12-ARIMA Nights spent: calendar adjusted time series 25,, 2,, 15,, 1,, 5,, Jan-9 Jan-91 Jan-92 Jan-93 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan- Jan-1 Jan-2 Jan-3 Jan-4 Jan-5 Jan-6 Jan-8 calendar adjusted for comparisons with the previous year S T A T I S T I C S A U S T R I A 12 6

7 Method X12-ARIMA Nights spent: calendar adjusted time series 25,, 2,, 15,, 1,, 5,, Jan-9 Jan-91 Jan-92 Jan-93 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan- Jan-1 Jan-2 Jan-3 Jan-4 Jan-5 Jan-6 Jan-8 calendar adjusted Easter + 2 weeks 4 days in March 12 days in April S T A T I S T I C S A U S T R I A 13 Method X12-ARIMA Nights spent: calendar adjustment - example 16,, 14,, 12,, 1,, 8,, 6,, 4,, 2,, Mar-6 Apr-6 Mar-7 Apr-7 Mar-8 Apr-8 calendar adjusted 26: Easter April 16 th 16 days in April 27: Easter April 8 th 15 days in April 28: Easter March 23 th 16 days in March adjust to 4 days in March and 12 days in April S T A T I S T I C S A U S T R I A 14 7

8 Method X12-ARIMA Nights spent: calendar adjustment - example Period Original data % change to same month in previous year Calendar adjusted time series % change to same month in previous year Jan-8 13,78, ,78, Feb-8 16,891, ,891, Mar-8 14,817, ,375, Apr-8 4,911, ,392, May-8 7,19, ,114, Jun-8 8,531, ,531, Jul-8 15,255, ,255, S T A T I S T I C S A U S T R I A 15 Method X12-ARIMA Nights spent: season and calendar adjusted time series 25,, 2,, 15,, 1,, 5,, Jan-9 Jan-91 Jan-92 Jan-93 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan- Jan-1 Jan-2 Jan-3 Jan-4 Jan-5 Jan-6 Jan-8 season and calendar adjusted for comparisons with the previous month S T A T I S T I C S A U S T R I A 16 8

9 Method X12-ARIMA Nights spent: season and calendar adjusted time series - example Period Original data % change to previous month Season and calendar adjusted time series % change to previous month Jan-8 13,78,35 1,286,394 Feb-8 16,891, ,68, Mar-8 14,817, ,463, Apr-8 4,911, ,951, May-8 7,19, ,85, Jun-8 8,531, ,116, Jul-8 15,255, ,574, S T A T I S T I C S A U S T R I A 17 Method X12-ARIMA Nights spent: trend 25,, 2,, 15,, 1,, 5,, Jan-9 Jan-91 Jan-92 Jan-93 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan- Jan-1 Jan-2 Jan-3 Jan-4 Jan-5 Jan-6 Jan-8 trend S T A T I S T I C S A U S T R I A 18 9

10 Problems & conclusion Problems Technical challenges Adjusted series contains the irregular component Tourism demand influenced by many factors (e.g. weather, marketing, foreign holidays, mega events, terrorist attacks, diseases, currency rate fluctuations, ) Tourist consumption can respond relatively easily and quickly Conclusion Time spent understanding the seasonal adjustment procedure is time well spent! S T A T I S T I C S A U S T R I A 19 Guglgasse 13 A-111 Vienna Tel.: +43 () johanna.ostertag@statistik.gv.at S T A T I S T I C S A U S T R I A 1