| This dissertation addresses the question of why the expectations hypothesis which postulates that the forward rates are unbiased estimators of future spot rates, is rejected by almost every empirical study. Explanations given for these rejections include: market inefficiency, overreaction of the long rates, measurement errors, and time varying term premia. In this study, we investigate the time varying term premia and measurement error explanations.; The first part of the dissertation focuses on the forecasting power of forward rates. Past studies have documented that 1-month forward rates partially predict future 1-month spot rates while 3-month forward rates have no predictive power at all. Another study also finds 1-year forward rates have forecast power and the forecast power increases with the forecast horizon. In this study, we find 1-week forward rates have substantial predictive power for the path of future 1-week spot rates. This study explains this puzzling "U" shaped forecasting power as measured by the adjusted R-square by examining the 3 components of the adjusted R-square, the expected change in rate, time varying term premium, and the forecast error.; The second part of this study focuses on the time varying term premia and its impact on the expectations theory. Traditional theories of the term premia, for example Hicks' (1947) liquidity preference theory, generally maintain that the term premium is basically a risk premium due to risk aversion. Assuming this theory is true, we would expect the term premia to vary slowly over time because individuals' attitudes toward risks change slowly. Hence a possible method of modelling the ex-ante term premia is to use past term premia. This study explores whether using past term premia as a proxy for the ex-ante term premia will increase the forecasting power of forward rates.; Finally, this dissertation addresses the measurement error problem in term structure research. Unlike past studies that focus on the measurement error in long rates, this study focuses on the effect of measurement error in the short rates on the forecasting power of forward rates. |
