| dc.contributor | 經濟系 | |
| dc.creator (作者) | 徐士勛 | |
| dc.date (日期) | 2018-10 | |
| dc.date.accessioned | 16-Jul-2025 11:13:07 (UTC+8) | - |
| dc.date.available | 16-Jul-2025 11:13:07 (UTC+8) | - |
| dc.date.issued (上傳時間) | 16-Jul-2025 11:13:07 (UTC+8) | - |
| dc.identifier.uri (URI) | https://nccur.lib.nccu.edu.tw/handle/140.119/158060 | - |
| dc.description.abstract (摘要) | 由於「理性預期」在經濟與財務理論中扮演一定的角色，因此當期經濟（或財務）變 數的實現值不但可能受到過去歷史資料的影響（此即「後顧式資訊」)，也可能反映對未來 各式衝撃或相關變數的預期（此即「前瞻式訊息」)。因此，如何透過計量模型將變數的時 間序列資料中這些能反映未來預期的「前瞻式訊息」給淬取出來，是此計畫所要硏究的 第一個議題。更進一步地，在一個體系當中，如何衡量各個變數對應的「前瞻式訊息」間 之相互影響或傳導關係（此即「外溢效果」)，進而建立對應的「鏈結網絡」，則是此計畫 最終要達成的目標。然而，當今文獻上所慣用的「自我相關迴歸模型」多建立在變數僅由 過去各期實現值與當期干擾項所決定的假設上，對於該時間序列中所可能蘊含的「前瞻 式訊息」則沒有直接對應的刻畫。可以想見，此類僅以「後顧式資訊」爲主的模型假設侷 限了可能的分析範疇而進而可能導致偏誤的推論。相較於文獻上已知的網絡硏究，我們 於此三年計畫中針對「前瞻式訊息」淬取以及對應的「鏈結網絡」建構，提出了一個較爲 全面且具可行性的分析架構。首先，我們以允許「後顧式資訊」與「前瞻式訊息」並存的 「自我相關迴歸模型」爲基礎，透過推導轉換將各變數對應的「前瞻式訊息」淬取出來。之 後，我們再據以建立「前瞻式訊息」對應之「向量自我迴歸模型」，並根據「葛蘭傑因果 關係檢定（Granger-causality Tests)」以及「廣義衝撃反應函數（Generalized Impulse Response Function)」的分析結果衡量一體系內「前瞻式訊息」間的彼此鏈結程度，進而 建構對應的「鏈結網絡」。透過各變數「前瞻式訊息」的「鏈結網絡」建立並與文獻上的已 知網絡分析比較，將有助於我們更進一步瞭解並預測此一體系中各變數的變化態勢與交 互影響關係。在此三年計畫中，我們除了完備分析架構細節與理論性質的確認，並討論可 能的擴充外，相關的實證議題應用也都將逐步進行並完成。就我們所知，此反映了一體系 中各變數「前瞻式訊息」的「鏈結網絡」分析架構與概念在相關文獻上仍未被提出，因此 預期此硏究成果將具有一定的參考價值與貢獻。 | |
| dc.description.abstract (摘要) | Because rational expectations about future relevant variables play a key role in many economic or financial theories, the variable of interest may be affected by not only its trend of past values but also expected future impulses, these backward-looking and froward-looking information are possibly all involved in determining the current value of the variable. How to extract the unobserved forward-looking component of a variable by econometric models/methods is of our interest. Besides, for these extracted forward-looking information of variables in a system, how to detect and measure their potential interdependence and to construct corresponding connectedness (spillover) network are the main goals of this project. However, either the linear autoregressive (AR) or Vector AR models commonly used in the literature, the temporal dependence in the process(es) is typically restricted to the past only, while following the standard Box-Jenkins methodology. These causal type of models do not directly take underlying forward-looking information into account, and thus may mislead the inference. Going beyond the studies on network analysis in the literature, for a system of variables of interest, this three-year project aims at proposing a new framework of establishing forward-looking connectedness network. To this end, we first extract the forward-looking component of each variable in this system based on its own mixed causal and noncausal AR model, then the corresponding VAR model is established for these extracted forward-looking components. Two types of connectedness measures (one is based on Granger-causality tests and the other is on generalized impulse response analysis) and thus the network of these forward-looking processes are thus constructed accordingly. Based on the proposed, many interesting empirical applications or issues will also be handled in this project. To our best knowledge, this project is the first attempt constructing the connectedness network for the unobserved forward-looking components of variables in a system, and will thus contribute to the related literature. | |
| dc.format.extent | 116 bytes | - |
| dc.format.mimetype | text/html | - |
| dc.relation (關聯) | 科技部, MOST106-2410-H004-010, 106.08-107.07 | |
| dc.subject (關鍵詞) | 理性預期; 前瞻式訊息; 外溢效果; 鏈結網絡; 前瞻式自我相關迴歸模型; 向量自 我迴歸模型; 葛蘭傑因果關係檢定; 廣義衝撃反應函數 | |
| dc.subject (關鍵詞) | Rational Expectation; Forward-looking Information; Spillover; Connectedness Network; Non-causal AR Model; VAR; Granger-causality Test; Generalized Impulse Response Function | |
| dc.title (題名) | 建構多變量序列所蘊含的前瞻性訊息鏈結網絡 | |
| dc.title (題名) | Constructing Forward-Looking Connectedness Network for a System of Multiple Processes | |
| dc.type (資料類型) | report | |
