利用R/S分析方法探討台灣匯率市場

Abstract

台灣屬小國且主要依賴於出口，因此匯率對台灣企業而言尤其重要，些微波動都可能造成公司龐大的匯損，本篇論文在探討台灣匯率市場及試圖找出預警工具，我採用R/S分析方法來求得赫斯特指數，它是用來衡量一個時間序列資料是否存在長記憶性的現象。本篇樣本資料為USDTWD，時間取自2000.1.3到2016.12.30，利用日報酬及月報酬資料來做分析，我發現台灣匯率市場不符合效率市場假說而是具有長記憶性，且其非週期性循環為60天及300天。在分析前，需先將資料取AR(1)的殘差項做為新的時間序列資料，再執行R/S分析，是為了消弭資料短期記憶性對赫斯特指數所造成的影響。將樣本分段做分析時，發現2009-2012這段區間，USDTWD存在正向持續性，自從2008金融海嘯後，聯準會實行量化寬鬆，導致有諸多國際熱錢流入台灣，而導致台幣不斷升值，這與分析結果相互對應，而在2013-2016期間，存在較多市場波動，而分析後也發現匯率並不存在顯著的正向持續性。最後，則建構出適當的局部赫斯特指數求算方式，來探討其對匯率大幅貶值的預警效果，發現指數雖具有預警效果，但其訊號出現到實際的匯率崩跌，仍存在約兩個月至半年的時間，故此工具不適合用在短期投資之參考工具，而適用於長期投資，且當赫斯特指數於半年內下滑的幅度超過0.1，則可視為匯率即將狂貶的前兆。

Taiwan is mainly dependent on exports, so the exchange rate is particularly important for Taiwan, and some currency fluctuations may cause the companies a huge loss. This paper discusses the USDTWD and attempts to find early warning tools. Then, we use the R/S analysis method to obtain the Hurst exponent, which is used to measure whether a time series data has long memory. Our sample is the USDTWD, from 2000.1.3 to 2016.12.30, using the daily and monthly returns. We find that the Taiwan exchange rate market does not meet the efficiency market hypothesis but has a long memory, and its non-periodic cycles are 60 days and 300 days. Besides, before the analysis, to eliminate the short-term data on the impact of the Hurst exponent, it is necessary to take the AR(1) residual term as the new time series data and then to carry out the R/S analysis. During the 2009-2012, it is found that USDTWD has positive persistence. Since the financial tsunami happened in 2008, FED implemented QE, resulting in a lot of money into Taiwan, which led to the appreciation of TWD, which corresponds to the analysis results. And during 2013-2016, there is more market fluctuations, and after the analysis, we find that USDTWD does not exist significant positive persistence. Finally, the appropriate method of acquiring the local Hurst index is constructed and to explore its early warning effect of the depreciation of the exchange rate. We find that the index has an early warning effect, but the signal appears to the abrupt depreciation, existing about 1-month to 10-month difference. So, the tool is suitable for long-term investment as the reference tool. And when the Hurst index falls more than 0.1 in a month or more than 0.15 in 2 months, it could be seen as a precursor to the large depreciation of the exchange rate.