Difficulties in Seasonal Adjustment N. Alpay KOÇAK Turkish Statistical Institute
Difficulties in seasonal adjustment In general manner, seasonal adjusment is a type of economic time series analysis . Trend-Cycle, Seasonal, Transitory, Irregular, Calendar Effect, Outliers, Missing Observations, Forecasts etc. Actually, one can use one of these components to interprete the properties of the economic time series.
Difficulties in seasonal adjustment Original series – seasonality = Seasonal Adjusted series (for short term economic analysis) Trend-cycle component (for short-term economic analysis with smoother series) Seasonal and calendar component (to understand underlying movements of the series in a year) Outliers (to understand shocks and persistent effects on the series) Etc...
Classical Output of Demetra+
Trend_Cycle = Trend + Cycle
Seasonal Component
Irregular component First shaded area represents the period between 2004 q2 – 2005q1 The second is 2008q3- 2009q2
Trading Days (Fixed holidays included)) Parameter Value Std error T-Stat P-value Monday -0.0187 0.0105 -1.77 0.0861 Tuesday 0.0007 0.0097 0.07 0.9447 Wednesday -0.0103 0.0111 -0.93 0.3611 Thursday 0.0286 0.0097 2.94 0.0061 Friday -0.0126 0.0106 -1.19 0.2424 Saturday 0.0050 0.0103 0.49 0.6309 Sunday(derived)0.0072 0.0091 0.79 0.4334 When the numbers of the Wednesdays are increased in a quarter, the volume of the series will increase in that quarter.
Difficulties in seasonal adjustment Two main requests for performing good seasonal adjustment of a time series Conceptual knowledge Statistical and Econometric knowledge
Difficulties in seasonal adjustment Conceptual knowledge Analytical Framework, Concepts, Definitions, and Classifications Definition Classification Scope of the data Methodologic issues Geographical issues Accounting Conventions Unit Valuation Compilation Practices We have to check that any changes in the definitons or classifications... Should be cleared the strucuture of the series. For wxample, National accoutns by Production approaches. If so, direct and indirect issues can be discussed. Value, rounding?
Statistical and Econometric knowledge Basic data transformations Model estimation (ARIMA model) Significancy Diagnostics Spectrum graphics Filters
Transformation: Level & Log
Statistical and Econometric knowledge Basic data transformations Model estimation (ARIMA model) Significancy Diagnostics Spectrum graphics Filters Not so much theoretical detail! Just to know Why and what they do?