台灣期貨交易所在2008年1月28日推出新台幣計價之黃金期貨，新台幣黃金期貨不但保持美元黃金期貨之特色並且消除了匯率風險、投資門檻較高等問題，希望提供一個優良的投資兼避險的管道。本篇研究以新台幣黃金期貨作為研究對象，並且使用濾嘴法則(Filter rule)、隨機指標(KD)、買入持有三種方法作為避險策略比較分析三種策略之獲利性差異與避險績效。新台幣黃金期貨價格之資料來源為AREMOS資料庫，現貨牌價資料來源則是台灣地區銀樓公會公告之買進及賣出價格之平均值作為現貨價格。本文之模擬操作日期為2008年1月28日至2010年12月31日，共有728筆日資料，並且將研究期間依現貨價格波動幅度區分為價格盤整期與價格上升期。本研究使用濾嘴法則、隨機指標以及買入持有三種策略。另外，再依照先前所區分的價格盤整期與價格上升期以及全面避險和最適避險分別進行三種策略的模擬操作，分別依不同的避險期間(3個月、6個月、12個月)對其績效做比較分析。研究顯示，濾嘴法則以濾嘴0.5%之績效最佳，不同持有期間以三個月之績效最佳或是說虧損最小。KD指標在不同持有期間下皆為正報酬且持有期間越長，其報酬越大。利用最小平方法求出最適避險比率後再進一步算出之最適避險績效與全面避險之績效差異並不大。三種策略相互比較後，以KD指標的表現最佳。 TAIFEX NT Dollar Gold Futures was launched on January 28, 2008 by the Taiwan Futures Exchange. TAIFEX NT Dollar Gold Futures not only maintains the characteristics of the TAIFEX US Dollar Gold Futures, but also eliminates of exchange rate risk and reduces investment threshold. It provides an excellent investment method for investors. This research uses TAIFEX NT Dollar Gold Futures as the research object. The Filter rule, Stochastic index and buy-and-hold three methods are used to do comparative analysis of the profitability differences and hedging performance of the three hedging strategy.Data of futures prices and spot prices were from AREMOS database and the announcement of Taiwan Jewel Association, respectively. Data were collected from January 28, 2008 to December 31, 2010, total of 728 daily data were included in the study. According to the movement of spot prices, study period was divided into two separate periods, which were to fluctuation period and rise period.In this research, we found that the smaller the filter is the higher profit investors can get, in general. However, 2% filter for 12 months would generate the best profits during the price fluctuation period, 0.5% filter for 6 months would have the least losses. While using Stochastic index as the tool of analysis, the longer investor holds the product, the more profit he would receive. During whole period buy-and-hold performed the best. If compare the Filter rule and Stochastic index, Stochastic index would have better performance.
