從1988年至今，已有相當大量的實證文章探討利用期間結構利差來做為預測未來實質GDP成長率的重要變數。學者們發現，當長短期債券之間利差愈大時，即將在未來的實質產出成長率會愈高；而當長短期債之間的利差愈小時，未來的實質產出成長率會愈低。這個運用利差作為變數的預測模型用於預測未來一至兩年間之實質產出累積成長率，表現尤其不錯。 然而，近期的文獻中指出，運用這個利差模型來預測未來實質產出，其預測能力有下降的趨勢，而最直接的證據便是這個預測模型對於1990到1991這段期間發生的蕭條幾乎沒有警覺到。 所以本文延伸這個以利差為重要解釋變數之實質GDP成長預測模型，並使用美國相關的總體資料來做實證研究（長期債利率使用十年期國庫券；短期債利率使用三個月期國庫券），將會在模型中逐步加入除了利差之外的其它重要解釋變數，以提高整體模型的解釋能力。接著並進一步的論證在考慮了1982年Greenspan明令貨幣政策工具改變這個事件後，對整個實質產出成長預測模型的影響。 最後，本文導入理性預期的觀念與思考邏輯，在透過思考不同的貨幣政策工具所提供市場的資訊多寡下，運用反向思考找到了一個雖簡單但對於整個模型預測能力有顯著提升的方法。 From 1988 to the present, a large empirical literature has used term structure yield spread that help predict future real GDP growth. This available empirical evidence tell us that high spreads today are associated with higher GDP growth in the future and low spreads today are associated with lower GDP growth in the future. This predicted model is used for predicting the GDP growth following one to two years especially good. However, the more recent evidence tells us that yield spreads have become less useful as predictors in recent years. In fact, one major predictive failures of the spread occurred on the occasion of the 1990－1991 recession. So in this paper, we use relevant macroeconomic data of USA to confirm and extend the conclusion of earlier studies. We discuss the yield spread between the 10-year Treasury bond rate and the 3-month Treasury bill rate contains information about future real GDP beyond that contained in various measures of monetary policy or inflation rates. Then we demonstrate that the ability of the yield spread to predict future output fluctuations is contingent on the change of monetary authority’s policy tool. Finally, we channel into the idea of rational expectation to consider different number of quantity of information in different case and find a simple but useful way to raise the predictive power of this model.
