經濟衰退一直是各國政府關心的議題，如果能夠事先預測到衰退發生的時間，貨幣當局才能透過調整貨幣政策工具，以避免衰退的發生。自1980年代起學者們陸續發現期間利差能夠顯著預測未來經濟活動，實證結果指出期間利差的預測內涵，不僅僅是含有短期貨幣政策的訊息，還包含巿場參與者對未來經濟情勢的預期。 本文引用Estrella and Hardouvelis(1991)提出之期間利差預測衰退的模型，以美國1960年第一季至2007年第一季樣本資料，探討期間利差預測衰退的能力。其中被解釋變數為美國經濟研究局（National Bureau of Economic Research）所公布之衰退時點，期間利差為十年期公債殖利率與三個月國庫券利率差距。接著分別於模型中加入聯邦基金利率（Federal Funds Rate）及通貨膨脹率，探討對模型預測能力的影響，本文實證結果再度證實期間利差模型極佳的預測能力。 然而，期間利差模型預測1990年-1991年景氣衰退失靈，使得相關學者們重新檢視期間利差模型的內涵，相關研究顯示期間利差模型預測經濟衰退仍有其顯著的表現。他們並且注意到1980年代起貨幣反應函數的改變，可能足以解釋期間利差模型預測能力下降的原因。 本文接下來導入理性預期觀點，探討貨幣政策中無法預期的部分是否會影響期間利差模型的預測能力，以利驗證理性預期假說之正確性。實證結果顯示利用理性預期觀點得出之模型，其模型解釋能力的確較一般期間利差模型解釋能力佳，顯示理性預期假說之頑固性。 Recession is an import issue for the government . If a predictor provides significant signals prior to recessions, the monetary authority can use their operating instruments to prevent the episode of recessions. From the 1980s , research has shown that term spread help predict future real economic activity in the United States and in some other industrial countries. Researchers have offered two reasons for this empirical relationship. First , term spread may contain information about the short-term monetary policy. The other reason is that term spread reflects market expectations of future economic activity. In this paper, we follow the paper of Estrella and Hardouvelis（1991）to predict recession. We use quarterly data from 1960:Q1 to 2007:Q1 in the United States to confirm and extend the conclusion . We use the data of recession dated by NBER as explained variable and the spread between the rate on the 10-year government bond and 3-month treasury bill rate as explanatory variable. Then we add the other independent variables such as federal funds rate and inflation rate. We want to examine the predictive content of model. The empirical result of the paper tells us that term spread also has prominent predictive power. However, the more recent evidence shows that term spread have become less useful in recent years. In fact, term spread model failed to predict the episode of the 1990－1991 recession. Hence, many researchers reexamine the predictive content of term spread . The result is consistent with former research. The change of Federal reaction function from the 1980s may help explain the reason of declining predictive power of term spread model. Finally, we channel into the idea of rational expectation to consider if the unexpected part of monetary policy could affect the predictive power of model. Furthermore, we test the robustness of rational expectation. Our research shows that term spread model bases on the theory of rational expectation is better than term spread model.
