Abstract: | 對於序列資料而言, 非定態 (non-stationarity) 是其中一個相當重要的特性, 其 存在將可能使常用的分析方法產生偏誤的推論。 為了增加對非定態特性檢定的檢定力 (testing power), 近年來文獻上利用不合季節性因素的追蹤資料 (panel data), 探討單 根 (unit root) 與共積 (cointegration) 關係的研究已經有很大的進展。 相對地, 文獻上 對於追蹤資料中可能真季節性的非定態特性的研究卻相當有限。 由於季節性的非定態 特性是男一個資料中常見的重要特性, 因此這一個兩年計畫將以 「研究追蹤資料中的 真季節性之非定態特性J 作為主要標的。 相較於已知的文獻研究, 我在這個計畫中將延 伸 Bai and Ng (2004, Econometrica) 的模型架構, 進一步分析資料中這些真有季節 性的非定態特性的來源。 同時, 為了增加季節性單根 (seasonal unit roots) 的檢定力, 我也根據個別序列檢定的 p 值提出了相對應的組合檢定 (pooled tests)。 除此之外, 我 也將在這樣的分析架構下探討真有季節性之共積 (seasonal cointegration) 問題, 並將 分析架構進一步延伸處理可能的結構性改變 (strucural changes) 問題。 在目前關於季 節性非定態的追蹤資料研究的文獻中, 就我所瞭解, 並不存在這樣的研究分析架構。 再 者, 利用此計畫的研究架構與成果, 我將分析台灣未經季節調整的總體資料, 藉以釐清 其中真季節性之非定態特性。 Non-stationarity is one of most important features of a series, and may severely bias the conventional analysis. For increasing the testing power of non-stationarity, there have been great progress and sizeable literature on the unit roots and coin- tegration analysis in (non-seasonal) panels. While seasonal non-stationarity is also crucial to seasonal unadjusted data, the analysis is relatively limited instead. This two-year project thus focuses on seasonal non-stationarity in panels with cross- section dependence. However, unlike the existing works, I go beyond them to analyze the source of seasonal non-stationarity. I establish a factor model for sea- sonal panels by extending PANIC approach of Bai and Ng (2004, Econometrica) to the case where possible seasonal non-stationarity is permitted. To increase the testing power, I also introduce pooled tests by using p-values of individual tests. Besides, based on the proposed framework, we can draw inferences on seasonal cointegration, and extend it to dealing with possible structural changes in seasonal panels. To my best knowledge, for seasonal panel data, there is no such analysis yet in the literature. As for the empirical studies, I will examine Taiwan’s seasonal un-adjusted macro data based on the proposed approach. |