應用高頻資料估計移動節日效應－－臺灣貨幣總計數的實證分析

Using High Frequency Data toModel Moving Holiday Effects: An Empirical Investigation of TaiwaneseMonetary Aggregates

中國新年在臺灣是一個重要的節日，並且在經濟活動上具有相當程度的影響。新年假期的時點是由農曆決定， 使得每年農曆年節期間對照在陽曆時，會變成具有移動節日效果的現象，如此之下，在臺灣的部分經濟變數則將隨農曆年節的移動而改變。另一方面， 本文嘗試使用日資料建構月的節日迴歸量變數來分析固定季節因子中的移動季節因子。一般而言， 在無日資料可以使用之下， 都是假設在節日期間每日的影響程度都是一樣的。在分析的時候， 本文定義影響節日效應的變數分別為「節日前」、「節日期間」和「節日後」這三種。分析的資料則是以臺灣的貨幣供給量，分別比較在沒有節日變數下、用標準的節日變數和用日資料建構的節日變數這三種。在迴歸量建構的方式上， 考慮積存量迴歸量和流量迴歸量這二種迴歸量。實證結果顯示， 在控制移動節日效應方面， 使用日資料下可以更精確的描述節日的影響效應，但改進的幅度不是非常大。

Chinese New Year is the most important holiday in Taiwan and strongly affects economic activities. The holiday is based upon lunar calendar and becomes a moving holiday in Gregorian calendar. We propose to use daily data to construct monthly holiday variables to distinguish the moving holiday factors from the regular seasonal components. With the availability of daily data, one is exempted free from being forced to assuming equal effect for each day within the holiday interval. We define three holiday variables: “before”, “during”, and “after” the holiday effect. We analyze four monetary aggregates in Taiwan and compare the performance of three methods: no holiday variables, holiday variable without using daily data, and holiday variables using daily data. Two types of daily average holiday regressors are presented, one type modeled after flow holiday regressors and the other type modeled after stock holiday regressors. Empirical analysis confirms the importance of controlling for moving holiday effect and daily data does help improve the precision of estimating holiday effect though the margin is not large.