The Improvement And Application Of Spring Festival Model In X-12-ARIMA Seasonal Adjustment

Time series will be affected by many factors, the strong seasonal effects will not only cover up the real time information and fundamental changes which the series convey, but also make the non-seasonal characteristics of this set of data to be express out difficultly. Seasonal adjustment precisely targeting to eliminate seasonal effects in the sequence can be helpful for the changes in the time series to reflect the timely and accurate potential economic changes, so that data are comparable between different seasons. X-12-ARIMA model is a internationally widely used seasonal adjustment method based on filters, and while Spring Festival is one of the most important traditional festival, it has a profound impact on people’s spending habits, market behavior, as well as national policy. Currently the widely used spring festival methods include Univariate average-weights models, Multivariate average-weights models and Multivariate varing-weight models, based on these, this paper proposes a new Multivariate varing-weight models, the3-period exponent-moving weight spring festival model for both flow data and stock data.After that, total retail sales of consumer goods, consumer price index and money supply MO were seasonally adjusted successively. The results show that pre-holiday effect and holiday effect of spring festival on the total retail sales of consumer goods were significant, the after-holiday effect was not significant; all three effects of spring festival on the consumer price index were significant; as a stock data, the whole holiday effects of spring festival on money supply MO were also significant.Through comparative analysis of these different spring festival models, this study found that the proposed3-period exponent-moving weight model could effectively identify the spring festival effects, provide more information than the univariate models. Furthermore the3-period exponent-moving weight model has a similar role with the3-period linear-moving weight model on adjusting total retail sales of consumer goods and CPI, but has different estimated parameters. Considering money supply MO, the3-period exponent-moving weight model identified less outliers, had a more smart more smooth adjusted series than the3-period linear-moving weight model.Finally this study concludes that the proposed3-period exponent-moving weight model is not only a reasonable theoretical assumption, but also able to estimate the effects of spring festival on flow data and stock data. It is helpful for researchers to seasonally adjust Chinese economic series.