TECHNICAL NOTES ON THE SEASONAL ADJUSTMENT OF CPI
Consumer Price Index (CPI) provides a general measure of the changes in average retail prices of commodities bought by specific group of consumers in a given area and in a given period of time. It mainly measures the composite change in the retail prices of the various commodities over time.
The seasonal adjustment of a time series mainly refers to the isolation of seasonal fluctuations, leaving the basic trend of the observed series. Seasonal fluctuations can be due to composite effect of climates and institutional events which repeat more or less regularly each year. Specific factors that may affect the CPI include seasonality due to production cycles, demand due to school year or holidays, and practices such as increase in rental rates during the beginning of the year. After the removal of seasonal variations, the resulting series is referred to as the seasonally adjusted series or the deseasonalized series. By removing the effects of seasonality on the CPI series, analysis can be made on a month-on-month basis. Thus, seasonal adjustment allows comparisons over recent months and gives short-term trend movements for the series. In general, if seasonally adjusted CPI levels are lower than the unadjusted series, it means that seasonal factors push up prices relative to the expected trend.
III. SEASONALITY IN THE CPI
The over-all CPI is tabulated using six major commodity groups in the Philippines, Metro Manila (MM) and Areas Outside Metro Manila (AOMM). The six groups are: Food, Beverages and Tobacco; Clothing; Housing and Repairs; Fuel, Light and Water; Services; Miscellaneous. The last five groups listed comprise the non-food items. Initially, the CPI series for all items as well as the Food, Beverages and Tobacco (FBT), and non-food items for MM and AOMM were tested for the presence of seasonality assuming stability. In MM and AOMM, presence of stable seasonality were observed both for FBT and Non-food items. However, the CPI for all items indicated that there is no seasonality in the series mainly due to the exhibited opposite direction of peaks and troughs of the FBT and non-food items, thus, canceling out. Presence of stable seasonality though, was observed both for FBT and non-food items when analyzed separately. The current seasonally adjusted series are based on X11ARIMA88 built-in model (Multiplicative with log-transformation (0,1,1)(0,1,1) and Additive (0,1,1)(0,1,1)) fit to the 2001 data series.