An Empirical Analysis of Forward Rate Prediction Error and Time-Varying Term Premium

• Main Authors: Tsou Shih Ching, 鄒世菁
• Other Authors: Wu-Jen Zhuang
• Format: Others
• Language: zh-TW
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/65985185999225491205

碩士 === 淡江大學 === 財務金融學系 === 87 === Most empirical studies of the expectations hypothesis reject it. The rejection has generally been attributed to: (a) a time-varying expected term premium, or (b) expectation errors by market participants. In this paper I examine the time-varying expected term premium argument for the failure of the expectations hypothesis of the term structure of Taiwan CP rates. Furthermore, recent methodologies to measure time-varying premium are usually centered around regression equations. I develop a time-varying parameter model to estimate expected term premium, an unobserved component, embedded in forward interest rates. Then the behavior of the term premium is analyzed. We investigate the interest rate term structure with 30, 60, 90, 120, 150 and 180 days maturities. Using state-space model to estimate expected term premium from April 1991 to December 1998, a Kalman filter technique is used to obtain estimates of the expected term premium from forward rate prediction errors of one to five months ahead, one-through five-month interest rates. And maximum-likelihood techniques are available to estimate the model parameters. First, I find considerable variation in estimated premium over time. The autoregressive estimates indicate that premium show a certain degree of persistence over time. It reveals a strong positive relation between current and past premium. Results from estimation show that the term premium is important in explaining the behavior of the prediction errors, that is, this method is quite successful in capturing the essence of the time-series properties of premium terms. Second, the expectations theory is rejected for the longer forecast horizons, but cannot be rejected for the shorter maturity, such as 30 and 60 days rates in one month forecast horizon.