本文利用多元迴歸模型檢定各種金融危機前後,影響黃金價格因素是否存在顯著差異,同時也應用Logit模型預測黃金漲跌方向。實證結顯示:全樣本的估計結果顯示,黃金與石油價格成正向關係,而美元指數與黃金呈反向變化;與美國十年期公債殖利率卻呈反向關係,與恐慌指標成正向變動關係,與道瓊指數呈反向關係。而不同樣本期間結果顯示:黃金與油價的關係除了在次貸風暴期間為負號外,其他子樣本期間均呈正向關係;與十年期公債殖利率大多呈反向關係;與恐慌指數在2009年3月之前大多呈反向關係,之後卻呈正向關係。應用二分類的Logit模型預測黃金價格變動的方向,無論是全樣本或是各子樣本期間均發現,其正確率均在50%左右。意涵影響黃金價格的因素很多,要有非常顯著的正確率是相當不容易的。
This study employs multivariate regression method to examine if any factor affects gold price exited significantly different before and after various financial crises. in the meantime applies Logit model to predict the direction of gold price. This study draws the following conclusions about the relationship of gold for full sample: 1) positive relationship with crude oil price, 2) negative relationship with dollar index, 3) negative relationship with the 10 year bond yield, 4) positive relationship with VIX index, and 5) negative relationship with DOW index. As far as subsamples, this study demonstrates the following conclusions: 1) except subprime period while the crude oil is negatively related with gold, the others are positively related with gold price, 2) most of 10 years bond yield among subsample are negatively related with gold price, 3) most VIX index are negatively related with gold price among subsamples before March, 2009, after then are positively related with gold price, 4) through binary Logit model, both full sample and subsample found about 50% accuracy rate. This result implies there are so many factors exited to affect gold price. It is very difficult to have significant accuracy rate.