黃金具有保值、對抗通貨膨脹的特性,被視為熱門投資工具之一。建構黃金價格的計量模型,可幫助投資人評價與預測黃金價格,本研究以Enders and Siklos (2001) 的不對稱門檻共整合模型,檢視黃金價格與美國廣義貨幣供給、美元指數、美國消費者物價指數之間是否存在共整合,且是否長期維持穩定。並進一步使用門檻誤差修正模型(TECM),研究變數間短期動態關係,最後再對無共整合之期間實施Granger因果關係檢定。研究樣本採用月資料,期間取自2002年1月至2013年12月,而實證結果發現:2002年3月至2010年4月之間黃金與解釋變數間存在共整合關係,但自2010年5月後共整合關係不復存在。若以存在共整合關係期間之模型預測2010年5月至2013年12月的黃金價格,模型指出黃金價格在該時間存在被低估的現象而具有投資價值。
The Gold is an asset with inflation proofing, and regarded as a widely-held investment. Constructing a regression model can give more accurate evaluation and prediction on the gold prices. This study examines if there is a co-integration relationship among gold prices, US money supply, the dollar index and consumer price index by employing Enders and Siklos (2001)’s asymmetric co-integrate threshold model. This study aims to examine if the co-integration relationship was lasting during entire sample period. For the period with co-integration relationship, we use TECM model to estimate the speed of reversion to long-run equilibrium. For the period without the co-integration relationship, we employ Granger causality testing to actually show which series lead.This study uses monthly data ranging from January 2002 to December 2013, and shows that there is a relationship of co-integration during the period between March 2002 and April 2010. However, the phenomenon of co-integration disappears after May 2010. Using the model of co-integration period to forecast the gold prices during May 2010 to December 2013, this study identifies an investment opportunity which is accrued by undervalued gold prices.