A PCA based term structure model with time-varying term-premia on government bonds – University of Copenhagen

Department of Mathematical Sciences > Calendar > Events > A PCA based term struc...

A PCA based term structure model with time-varying term-premia on government bonds

Specialeforsvar ved Nina Tranegaard Berril Andersen

Titel: A PCA based term structure model with time-varying term-premia on government bonds

Abstract: The yield curve can be decomposed into three components: expected inflation, expec-tations about the future path of real short-term interest rates, and a term premium. The term premium is the extra return that lenders demand to hold a long-term bond instead of investing in a series of short-term securities. With yields trending lower and lower in recent years, the interest of understanding the behaviour of the term premium has in-creased. This thesis presents a PCA based term structure model from which we can extract the model-implied term premium. The model is estimated for US Treasury securities. We discuss different historical events' impact on the 10Y Treasury term premium. An event study of the impact of the Federal Reserve's quantitative easing programs on the 10Y Treasury yield is carried out. The estimates imply that the 10Y Treasury yield fell cumulatively by 194 basis points on days that quantitative easing measures were announced. Around 90\% of the decline in the 10Y Treasury yield around these announcements can be attributed to a decline in the 10Y Treasury term premium. We comment on the uncertainty regarding these estimates. A linear term premium model where the historical time variation in the 10Y Treasury term premium is explained by selected business cycle and uncertainty variables are estimated. The model estimates show that the term premium is a countercyclical variable which tends to move with measures of uncertainty and disagreement about the future level of yields. Applying the model to German yield data, we find that the 10Y government bond term premia in the US and in Germany are highly correlated

Vejleder:  David G. Skovmand
Censor:    Jesper Lund, CBS