Title page
Contents
Abstract / Résumé 4
Introduction 7
1. Is there seasonality in consumer prices? 9
2. How do NSOs and Central Banks adjust CPI for seasonality? 17
3. How well do standard seasonal adjustment methods perform in normal times? 19
4. Seasonal adjustment in times of large shocks 25
5. Experimental methods and seasonal adjustment 32
6. Communication and guidance to users 32
7. Conclusion 37
References 39
Annex A. Visual inspection of seasonality in headline CPI in non-G7 economies 42
Annex B. Standard tests and methods 48
Annex C. COICOP 99 categories used in the analysis 53
Annex D. Additional information on the simulation 54
Table 1. Combined test of CPI seasonality by COICOP categories 13
Table 2. Change in seasonal patterns from 1980 to 2023 15
Table 3. Methods employed by NSOs and Central Banks to seasonally adjust CPIs 18
Table 4. Number of selected outliers by time interval and specifications 27
Table 5. Changes in outliers with the addition of time series data 28
Table 6. Results of seasonal adjustment with manual adjustment for outliers 29
Table 7. Conceptually determining structural breaks in seasonal adjustment 30
Table 8. Practices of communicating seasonally adjusted data in selected OECD countries 33
Table 9. Summary of the assessment 37
Figure 1. Seasonality in headline CPI in G7 countries 9
Figure 2. CPI "clothing and footwear" in G7 economies 11
Figure 3. Changes in seasonal patterns from 1980 to 2023 16
Figure 4. Seasonal patterns in headline CPI at the start of the pandemic 17
Figure 5. Seasonal adjustment using X-13 versus seasonal adjustment performed by NSOs 20
Figure 6. Comparison of seasonality detection between X-13 and TRAMO-SEATS over different time series lengths 22
Figure 7. Direct versus indirect seasonal adjustment with X-13, deviation from aggregated CPI index 24
Figure 8. Differences between X13 and TRAMO-SEATS revisions in G7 countries CPI indices 31
Boxes
Box 1. Seasonal products in CPI 14
Box 2. Communication in selected OECD countries with respect to revisions and seasonally adjusted series 35
Annex Tables
Table A B.1. Default Specifications in JDemetra+ 51
Table A D.1. Presence of an Easter effect by country 56
Table A D.2. X-13 and TRAMO-SEATS performance using default specifications 57
Table A D.3. Comparison of initial time series transformation choices using X-13 and TRAMO-SEATS 57
Table A D.4. Comparison between first and second differenced ARIMA models using X-13 and TRAMO-SEATS 58
Table A D.5. Comparison of different Easter effect specifications using X-13 and TRAMO-SEATS 59
Table A D.6. Correlation between OECD and NSO seasonal adjustment (month-on-month series) 59
Annex Figures
Figure A B.1. Seasonality test 49
Figure A B.2. Standard seasonal adjustment methods 50
Figure A B.3. Three types of outliers (AO, TC, LS) 52
Figure A D.1. Visualisation of simulated time series 55
Figure A D.2. Direct vs. indirect SA indices (X-13) 60
Figure A D.3. Direct vs. indirect SA month-on-month change (X-13) 61
Figure A D.4. Direct vs. indirect SA year-on-year change (X-13) 62
Figure A D.5. Direct vs. indirect SA indices (TRAMO-SEATS) 63
Figure A D.6. Direct vs. indirect SA month-on-month change (TRAMO-SEATS) 64
Figure A D.7. Direct vs. indirect SA year-on-year change (TRAMO-SEATS) 65
Figure A D.8. Calculated All-items CPI and NSO All-items CPI comparison 66