A Comparative Study On The Method Of Seasonal Adjustment Based On Total Retail Sales

Data which is measured in unit of month or season in time series is susceptible to seasonal factors and appears seasonal periodic change, which is called seasonal components. The exist of seasonal components will cover up the fundamental features of data which is directly bound up with the trend analysis of current economic development, and cannot reflect the trend variation or cyclical period change of social economical phenomenon objectively, since monthly or seasonal time series include seasonal change factors. The so-called seasonal adjustments is abandoning the influence of season factor by resolving the economic time series to make the development tendency graphics of time series become as smooth as possible by irregular, then we can study the changes of economical phenomenon more precisely and objectively.Traditional seasonal adjustment eliminates volatility of series to make them show a sort of tendency mainly by moving averages and smoothing methods. This method is too simple and rough to consider the data's features. Later, the appearance of methods such as X-11-ARIMA,X-12-ARIMA,TRAMO/SEATS brings a new leap in the theoretical study and improved methods of seasonal adjustment. These methods are technically advanced and have a good effect when seasonally adjust the data. However, Chinese seasonal data has her unique characteristics, such as the uncertainty of the lunar holidays, the uncertainty of the length of the vacation and the adjustment of the Spring Festival factors. Though these questions have been studied, the methods are not thorough perfect and the adjustment methods are not in a consensus. Specific to the particular case of Chinese seasonal data, we should firstly take example by methods such as X-12-ARIMA,TRAMO/SEATS, these methods are rather mature in data smoothing and better in eliminating volatility, so they can be directly applied. As for the question of the change of the lunar holiday in calendar month, the dummy variable method is employed to quantitative analysis the effect of the lunar holiday. We analyze the effect of adding holidays by means of adding dummy variable in the model. We borrow ideas from the adjustment method of trade day to analyze the effect of length change. As for the adjustment of the Spring Festival factors, we alter the shortcomings which fix the influence scales of the Spring Festival, adjust its influence scales and even adjust it by using varied nonlinear weight according to the different distance from the Spring Festival. We adjust the Chinese seasonal data by the constructed method, and then analyze the variation trend of economic phenomenon. The adjusted data has good effect on eliminating seasonal volatility and better reflects phenomenal trend variation.