Forecasting the expected returns on bonds with increasing certainty is wanted from all rational investors in the fixed income markets. The potential for higher returns increase with the ability to forecast expected returns, through better trading payoffs and improved hedging and risk management. The expectations hypothesis was long prevailing in the academical litterature. It stated that the rational investor was expected to require zero or at least a constant excess return on bonds with long maturity over short maturity. This is equal to no time varying risk premiums. It is however reasonable for the rational investor to have time varying risk preferences based on the economic situation and outlook for the future, as described by Cochrane (1999). Thus, bonds with different maturity may be priced with different risk in an efficient market, and accordingly have time varying risk premiums. The expectations hypothesis has thus been rejected. This has been manifested through the classical studies of Fama and Bliss (1987) as well as Campbell and Shiller (1991). These studies modelled predictions of bond returns on specific maturities, with a R2 up to 18%. In a new and original approach, Cochrane and Piazzesi (2005) models a single-factor that predicts bond returns of any maturity, with a R2 up to 44%, more than doubled from the studies mentioned above. This is done on the same dataset as Fama and Bliss (1987) used and would be a big discovery within the field, if the model can be accepted across time and datasets. I test the model of Cochrane and Piazzesi (2005) based on the framework that these used originally, as well as new tests they have provided as response to critique of the model. So far, no other paper has rejected this model on all these dimensions. I use very well accepted data, and reject the model in every dimension tested. This paper is thus the rejection of the Cochrane and Piazzesi (2005) single-factor bond forecasting model.
|Educations||MSc in Applied Economics and Finance, (Graduate Programme) Final Thesis|
|Number of pages||109|